Combination of information from the parallel processing of different basic features (color, size, etc.) can be used to guide attention to targets defined by conjunctions of those features. Wolfe et al. (1990) argued that, although it was possible to guide attention to the conjunction of, for instance, color and orientation, it was not possible to guide attention to targets defined by conjunctions of two colors or two orientations. The present experiments demonstrate an exception to this rule. Although it is true that attention cannot be guided to a target that has red and green parts, attention can be guided to an item that can be described as a whole red target with a green part. In Experiments 1 and 2, we illustrate this point. In Experiments 3 and 4, we rule out some simple size-based accounts of this finding. In Experiments 5 and 6, we begin to explore the nature of these first steps toward dividing the visual scene into parts and wholes.