Whether attention to a local part of a visual display can prevent access to semantic information in form matching tasks with objects was studied. A first picture containing a line segment (the reference) was followed by 2 lateral objects also containing a line segment (a target and a distractor). Participants matched the line segments according to either their orientation or color. Effects of semantic information were assessed by manipulating the semantic relations among the pictures surrounding the reference, target, and distractor. Semantic information affected performance in the orientation matching task, but not in the color matching task. Results suggest the existence of separate selection mechanisms in vision. Selection of local colors for response purposes can be based on inhibition of the form pathway (eliminating semantic effects on matching). Selection within the form pathway can involve a bias toward global shape (the picture). Once attention is allocated to global shape associated semantic representations are activated and semantic effects on matching emerge.