Four experiments are described in which 1 visual object (the target) was selected from another (the distractor) according to its color (Experiments 1, 2, and 4) or its relative location (Experiment 3) and then was classified according to a simple geometric property. Object classification was signaled as fast as possible by a precision or power grip response, and this grip was either compatible or incompatible with either object. When targets were selected by color, target-compatible grip responses were facilitated, but distractor-compatible grip responses were impaired. When targets were selected by location, similar results were obtained for target-compatible grip responses, but not distractor-compatible grip responses. These data are explained in terms of the involvement of action codes in object-level selection.