When we look at a chair or a giraffe we cannot suppress a semantic interpretation of that image, although we need not name it (e.g., Smith & McGee, 1980). Given that classification of object images is mandatory, is it capacity free? Subjects attempted to detect the presence or absence of a target object, specified by basic-level name, in a 100-ms display of a nonscene (clock face) arrangement of one to six pictures of common objects. There was a sharp monotonic decrease in detectability as a function of the number of objects in the display, indicating that object detection under these conditions is an attention-demanding process. No benefit was observed for targets that were likely to co-occur with the distractors. This latter result is evidence against an account of the perceptual interference found for improbable objects in real-world scenes, which holds that the interference derives from an inventory listing of the objects without regard to their spatial relations.