Heart rate and six speech measures were evaluated using a manual tracking task under different workload demands. Following training, 17 male subjects performed three task trials: a difficult trial, with a $50 incentive for successful performance at a very demanding level; an easy trial, with a $2 incentive for successful performance at a simple level; and a baseline trial, in which there was physiological monitoring but no tracking performance. Subjects counted aloud during the trials. It was found that heart rate, speaking fundamental frequency (pitch), and vocal intensity (loudness) increased significantly with workload demands. Speaking rate showed a marginal increase, while vocal jitter and vocal shimmer did not show reliable changes. A derived speech measure, which statistically combined information from all other speech measures except shimmer, was also evaluated. It increased significantly with workload demands and was surprisingly robust in showing differences for individual subjects. It appears that speech analysis can provide practical workload information.