The sense of agency (SoA) represents the everyday experience of control over our actions and their outcomes. We posit a new framework that defines SoA as consisting of three main components: sense of control of self, sense of control of the environment, and the presence of a goal. Across five experiments, we test this framework by altering participants' SoA over their actions and outcomes by manipulating the predictability of each. Results suggest that both actions and outcomes affect participants' SoA. We also report, contrary to previous theoretical predictions, that unpredictable outcomes lead to the lowest SoA as compared to actions. Additionally, results from explicit measures suggest that participants do not discriminate between control over actions and outcomes and that this remains true regardless of experimental design or explicit agency question type. Taken together, these results suggest that both actions and outcomes are vital to the experience of control.

In two studies we investigated the effect of resumption practice following an interruption at the same step in a Computerized Physician Order Entry system (CPOE). The results of both studies showed that error rate decreased with increasing amounts of resumption practice. One reason people may have resumed more accurately following an interruption is improvement in a general resumption process. If true, we would expect that participants could be interrupted at any step in a task and show improved resumption with increased practice. Instead, our results suggest that repeatedly resuming from the same step likely produces associative priming between a specific task, interruption, and step. The associative priming allowed participants to resume more successfully with additional interruption practice, but only for that task-interruption-step triplet.

We describe two different cognitive process models of a wellknown experiment on social influence (Salganik, Dodds, &Watts, 2006). One model, the social influence model,reproduced the choices that participants took by modelingboth the cognitive processes the participant engaged in andthe social influences that the participant saw. The secondmodel, the pure cognitive model, used only cognitivecapabilities and did not model any social influences that theparticipant saw. Somewhat surprisingly, the two modelsshowed no difference in quality of fit (the pure cognitivemodel actually fit slightly better than the social influencemodel), suggesting that social influence models should takecognitive functions into account in their theories.