Practice testing (i.e., practice retrieval) has been established as an effective learning strategy. Uncovering potential factors influencing self-testing usage is a prerequisite to promote its practical use. The present study reports five experiments exploring whether test anxiety (TA) and test stake (1) affect self-testing usage (Experiments 1-5) and (2) influence learning performance through their negative effects on self-testing usage (Experiments 1, 4, and 5). Experiment 1 analyzed data from 459 high school students collected via a survey and found both that TA negatively predicted students' daily use of self-testing and that self-testing usage mediated the negative association between TA and academic performance. The negative association between TA and self-testing usage was further replicated in a laboratory experiment (Experiment 2). Another quasi-experiment (Experiment 3) showed that students were less likely to test themselves when preparing for a high-stake than a low-stake exam. Experiment 4 replicated this finding and additionally demonstrated that a high-stake test led to poorer learning via its negative influence on self-testing usage. Experiment 5 demonstrated that a high-stake test provoked high state anxiety, which then induced avoidance of self-testing and ultimately impaired learning. Overall, these findings demonstrate a negative effect of TA on self-testing usage, in turn leading to poor learning. Practical implications are discussed.