We present the theoretical background and experimental validation of a pupil tracking method based on measurement of the irradiance centroid of Hartmann-Shack aberrometric images. The experimental setup consists of a Hartmann-Shack (HS) sensor forming over the same camera the images of the eye's pupil and the aberrometric image. The calibration is made by comparing the controlled displacements induced to an artificial eye with the displacements estimated from the centroid of the pupil and of the HS focal plane. The pupil image is also used for validation of the method when operating with human eyes. The experimental results after calibration show a root mean square error of 10.45 mum for the artificial eye and 27, 10, and 6 mum rms for human eyes tested using Hartmann-Shack images, with signal-to-noise ratios of 6, 8, and 11, respectively. The performance of the method is similar to conventional commercial eye trackers. It avoids the need for using separate tracking devices and their associated synchronization problems. This technique can also be used to reprocess present and stored sets of Hartmann-Shack aberrometric images to estimate the ocular movements that occurred during the measurement runs, and, if convenient, to correct the measured aberrations from their influence.