A major obstacle in point-of-regard monitoring for human-computer interaction has been the contami-Keywords: Human-computer interaction; Point-ofregard; Sensor; Eye model; Multi-layer perceptron; Head movement artefact as well as in the area discussed here of humancomputer interaction.Methods of eye movement measurement vary widely, not least regarding accuracy, convenience and cost. The highest achievable accuracies are unnecessary in determining the point-of-regard, since there exists a discrepancy typically of up to 1 degree between this and the physical direction of the eye [2]. This uncertainty is due mainly to the size of the fovea (that is, the small central region of the retina possessing highest visual acuity). Therefore, a sensor accuracy of between 0.25 and 2 degrees is usually considered sufficient and appropriate in this application [3][4][5][6][7]. This allows a beneficial trade-off to be made between sensor complexity and accuracy. In estimating a seated individual's point-of-regard on a computer display a universal problem to overcome is the contaminating effect of head movements. This paper addresses the problem of head movement artefact and presents a solution which has several novel features that make it potentially convenient to use and simple to produce.