His presentation describes experiments that explored eye behaviour when carrying out purely visual tasks on a Corel database of 1000 images. Results are reported that indicate that image identification can be carried out significantly faster with an eye tracker than with a mouse. Participants performing image search tasks were able to reach target images in significantly fewer steps with an eye tracker than by random selection. The effects of the intrinsic difficulty of finding images and the time allowed for successive selections were also investigated. Finally the results yielded evidence of the use of rapid pre-attentive vision during visual search.