Are photographs of objects presented on a screen in an experimental context treated as the objects themselves or are they interpreted as symbols standing for objects? We addressed this question by investigating the size Stroop effect—the finding that people take longer to judge the relative size of two pictures when the real-world size of the depicted objects is incongruent with their display size. In Experiment 1, we replicated the size Stroop effect with new stimuli pairs (e.g., a zebra and a watermelon). In Experiment 2, we replaced the large objects in Experiment 1 with small toy objects that usually stand for them (e.g., a toy zebra), and found that the Stroop effect was driven by what the toys stood for, not by the toys themselves. In Experiment 3, we showed that the association between an image of a toy and the object the toy typically stands for is not automatic: when toys were pitted against the objects they typically represent (e.g., a toy zebra versus a zebra), images of toys were interpreted as representations of small objects, unlike in Experiment 2. We argue that participants interpret images as discourse-bound symbols and automatically compute what the images stand for in the discourse context of the experimental situation.