User performance with a tabletop display was tested using touch-based and mouse-based interaction in a traditional pointing task. Dependent variables were throughput, error rate, and movement time. In a study with 12 participants, touch had a higher throughput with average of 5.53 bps compared to 3.83 bps for the mouse. Touch also had a lower movement time on average, with block means ranging from 403 ms to 1051 ms vs. 607 ms to 1323 ms with the mouse. Error rates were lower for the mouse at 2.1%, compared to 9.8% for touch. The high error rates using touch were attributed to problems in selecting small targets with the finger. It is argued that, overall, touch input is a preferred and efficient input technique for tabletop displays, but that more research is needed to improve touch selection of small targets.