Although students tend to dislike exams, tests—broadly defined in the present commentary as opportunities to practice retrieving to-be-learned information—can function as one of the most powerful learning tools. However, tests have a variety of attributes that affect their efficacy as a learning tool. For example, tests can have high and low stakes (i.e., the proportion of a student’s grade the exam is worth), vary in frequency, cover different ranges of course content (e.g., cumulative versus non-cumulative exams), appear in many forms (e.g., multiple-choice versus short answer), and occur before or after the presentation of what is to be learned. In this commentary, we discuss how these different approaches to test design can impact the ability of tests to enhance learning and how their use as instruments of learning—not just means of assessment—can benefit long-term learning. We suggest that instructors use frequent, low-stakes, cumulative exams and a variety of test formats (e.g., cued recall, multiple-choice, and true/false) and give students exams both prior to learning and following the presentation of the to-be-learned material.