According to the concept of desirable difficulties, introducing difficulties in learning may sacrifice short-term performance in order to benefit long-term retention of learning. We describe three types of desirable difficulty effects: testing, generation, and varied conditions of practice. The empirical literature indicates that desirable difficulty effects are not always obtained and we suggest that cognitive load theory may be used to explain many of these contradictory results. Many failures to obtain desirable difficulty effects may occur under conditions where working memory is already stressed due to the use of high element interactivity information. Under such conditions, the introduction of additional difficulties may be undesirable rather than desirable. Empirical evidence from diverse experiments is used to support this hypothesis.