This classroom experiment investigates the effects of adding representational pictures to multiple-choice and constructed-response test items to understand the role of the response format for the multimedia effect in testing. Participants were 575 fifth- and sixth-graders who answered 28 science test items--seven items in each of four experimental conditions in a balanced 2 × 2 within-subject design, with the factors being "multimedia" (text-only vs. text-picture) and "response format" (multiple-choice vs. constructed-response). Consistent with multimedia and generative learning theory, there was a multimedia effect for testing in which students were more successful in solving the items, gave higher metacognitive ratings of expected success, and gave higher satisfaction ratings for test items that contained text and corresponding pictures than text alone both for multiple-choice and constructed-response items. Consistent with problem solving theory, there was a response format effect in which students were more successful and gave higher metacognitive and satisfaction ratings for multiple-choice items than constructed-response items. Furthermore, pictures were slightly more beneficial for improving students' performance on constructed-response items as compared with multiple-choice items. Thus, the addition of pictures and multiple-choice response options to word problems can be employed to adapt item difficulty by design and support instructional control.