Learners typically allocate more resources to learning items that are higher in value than they do to items lower in value. For instance, when items vary in point value for learning, participants allocate more study time to the higher point items than they do to the lower point items. The current experiments extend this research to a context where the value of items is not explicitly labeled by evaluating whether learners' study decisions are sensitive to statistical regularities in the testing environment that signal which items are valuable to learn. In 4 experiments, participants studied English-English and Swahili-English paired associates across 4 study-test trials. On each trial, they were tested on only 1 type of item (e.g., only Swahili-English pairs), and, hence, only 1 type of item was valuable for performance. Some participants were cued to which information was valuable, and other participants had to learn from task experience. Experiment 2 examined the effect of performance incentives on study decisions, and Experiment 3 examined how the organization of the task environment influences learners' decisions. Finally, Experiment 4 examined the role of working memory span, fluid intelligence, and need for cognition on decisions. Findings indicated that some people can learn from task experience which items are important to study. However, many learners fail to do so (in particular, learners with low working memory spans); as a consequence, they dysregulate their study by allocating time to items that are not valuable to performance.