In the last fifteen years, there has been an explosion in the application of experimental cognitive paradigms within the research on emotional disorders. One of the most fruitful lines of research in this area is the study of selective attention in persons suffering from an emotional disorder. Several paradigms have been employed to examine this attentional bias (e.g. dichotic listening, Emotional Stroop, visual dot-probe task) and to demonstrate how task performance is facilitated or inhibited due to the presentation of a stimulus that is related to the emotional concerns of the participants. A problem associated with these paradigms is that they only measure attention deployment at a very specific moment immediately after the presentation of the emotional stimulus, but are not suitable for capturing the course of selective attention over longer time periods. In a study with spider anxious participants, we used on-line registration of eye movements as a continuous index of attention deployment towards emotionally relevant (spiders) or irrelevant (flowers) material. Viewing patterns were registered during a 3-second presentation of stimuli composed of a picture of a spider and a picture of a flower. Results show that spider anxious participants looked significantly more at spiders than at flowers during the beginning of the stimulus presentation, but subsequently their viewing pattern shifted more and more away from the spiders. Control participants showed a more stable pattern as they looked more at spiders than at flowers throughout the trial. A traditional attentional bias effect could, however, not be replicated.