Pictures used to supplement survey questions can systematically influence the answers obtained. Respondents react to the content of the image, giving higher-frequency reports when pictures of high-frequency events are shown and lower-frequency reports when pictures of low-frequency events are shown. The effects of pictures on responses are similar to those of verbal instructions (i.e., they produce an assimilation effect). Our results show that verbal and visual language both have independent effects as well as interact with each other. Verbal instructions have stronger effects than the visual effects produced by pictures, however, and can be used to counteract the visual context effects. We find that respondents pay more attention to verbal instruction when the verbal and visual cues are inconsistent with each other. This article provides evidence for a hierarchy of features that respondents attend to, with verbal language taking precedence over visual cues like pictures. Effective question writing, with verbal instructions making the question clear to respondents, reduces visual context effects. We found little evidence that conditions with pictures were evaluated better than conditions without pictures.