In this study, we present 3 picture-word interference (PWI) experiments designed to investigate whether lexical selection processes are competitive. We focus on semantic associative relations, which should interfere according to competitive models but not according to certain noncompetitive models. In a modified version of the PWI paradigm, distractor word pairs were simultaneously presented with the target picture. The distractor words were orthographically related directly to the target picture name (distractors: "came"ra bag"el"; target: camel), indirectly related to the name of a semantic associate of the target (distractors: "came"ra bag"el"; target: pyramid, an associate of camel), or unrelated. In a first experiment, which included only indirect relations, we failed to find interference from indirectly activated associates. However, in 2 subsequent experiments that included the associates as naming trials within the experiment, we demonstrated that indirect, orthographically mediated activation of associates produces reliable interference effects. The results indicate that semantic interference is not restricted to members of the same category and are problematic for models of lexical selection that do not include lexical competition. (Contains 1 table, 2 figures and 5 footnotes.)