Much of the learning that college students engage in today occurs in unsupervised settings, making effective self-regulated learning techniques of particular importance. We examined the impact of task difficulty and supervision on whether participants would follow written instructions to use repeated testing over restudying. In Study 1, we found that when supervised, instructions to test resulted in changes in the self-regulated learning behaviors such that participants tested more often than they studied, relative to participants who were unsupervised during learning. This was true regardless of the task difficulty. In Study 2, we showed that failure to shift study strategies in unsupervised learning was likely due to participants avoidance of testing rather than failure to read the instructions at all. Participants who tested more frequently remembered more words later regardless of supervision or whether or not they received instructions to test, replicating the well-established testing effect (e.g., Dunlosky et al. in "Psychol Sci Public Interest" 14:4-58, 2013. http://bibliotheek.ehb.be:2188/10.1177/1529100612453266). In sum, there was a benefit to testing, but instructing participants to test only increased their choice to test when they were supervised. We conclude that supervision has an impact on whether participants follow instructions to test.