Four experiments sought to disentangle the relative benefits of teaching expectancy (i.e., preparing to teach) and actually teaching (i.e., explaining to others) on learning. Participants studied a paper-based multimedia lesson on the Doppler Effect with the expectation of teaching or being tested on the material; some participants actually taught the material (by providing a video-recorded lecture), whereas others only studied the lesson. Participants then completed a comprehension test either immediately or following a one-week delay. Results indicated that when tested immediately (Experiments 1 and 3), those who prepared to teach outperformed those who prepared to be tested (i.e., a teaching expectancy effect; Experiment 1: d = .59; Experiment 3: d = .55); further, the act of teaching did not improve learning beyond preparing to teach. However, when tested following a one-week delay (Experiments 2 and 4), participants who actually taught the material outperformed those who only prepared to teach (i.e., a teaching effect; Experiment 2: d = .79; Experiment 4: d = .90), and the teaching expectancy effect was eliminated. Overall, these findings suggest that preparing to teach helped students better manage essential processing (i.e., processing related to initially representing the material), which may result in short-term benefits, whereas actually teaching fostered deeper generative processing (i.e., processing related to making sense out of the material), which is critical for achieving long-term learning.