What you have seen before helps you see it again. This effect has been shown in visual search studies looking at the consequence of the previous trial: Reaction times are shorter when the features defining a target and distractors are repeated. Here, I explore whether this bias in attentional selection occurs automatically or whether it depends upon the current goals of observers. Participants performed a visual search task, in which both a color singleton and a shape singleton appeared in the search array. The observers were instructed at the beginning of every trial as to which singleton was relevant. The data show that repeating the color or shape from the previous trial benefits performance only when this information is relevant to the observers' current goals.