The camera-robot calibration requirements in visual servoing can make it difficult to apply in many real-world situations. One approach to image-based visual servoing without calibration is to dynamically estimate the full visual-motor Jacobian during motion. In this paper we contrast three commonly used estimation approaches, and compare errors in the estimates. Other work has reported endpoint positioning accuracy for uncallibrated visual servoing. We also show how varying estimates errors affect trajectory following. The experiments have been carried out on both real and simulated PUMA 560 and 760 robots.