Mutual gaze is an important conversational resource, but is difficult to provide using conventional video conferencing equipment due to the disparity between the position of the camera and the position of the eyes on the screen. Various elaborate inventions have been proposed to get around this problem but none have found wide use. The alternative explored here is that these expensive alternatives may be unnecessary. Users of conventional desktop video equipment may, under the right conditions, be able to learn to interpret what is at first sight inappropriate apparent gaze direction as signalling that the other person is "looking at me."Data are presented from two experiments where an estimator judges where a gazer is looking. The gazer may be looking either at the desktop video image of the estimator or some point to the side. Experiment 1 compared two image sizes and two camera positions. While the size of the image (352 × 288 pixels versus 176 × 144) had no significant effect on participants' ability to judge where the gazer was looking, horizontally offsetting the position of the camera inhibited performance. Experiment 2 examined the effect of reducing the image size further. The smallest image size (88 × 72 pixels) resulted in poorer performance than the intermediate (176 × 144). The results show that it is possible for users of low cost desktop video conferencing to learn to interpret gaze direction to a very high degree of accuracy if the equipment is configured optimally. The practical and theoretical implications of these results are discussed.