As the use of survey experiments has spread throughout political science, experimental designs have grown increasingly complex. Yet, most survey experiments rest on a basic protocol by which treatments are delivered with textual vignettes and the effects of these interventions are then measured using self-reports of political attitudes or behaviors. We outline several design innovations that allow researchers to move beyond self-reports by directly embedding politically-relevant behaviors into survey experiments. As described in this chapter, these innovations enable experimentalists to strengthen the power of their treatments while enhancing the validity of their measures of treatment effects. We document these advances with illustrations drawn from a wide range of studies focusing on exposure to news reports, party polarization, racial prejudice, and physiological arousal.