In this study, we focused on the effect of reflection on different instructional methods, comparing observational learning and learning by doing, in the context of an academic writing task. Our goal was to investigate how reflection and instructional method affect academic writing performance, self-efficacy beliefs and students' satisfaction with the learning activities. In a quasi-experiment, 111 undergraduate students were assigned to either an observational learning or learning-by-doing condition, with or without reflection. In the observational learning condition students learned by observing a weak and strong model's writing processes. In the learning-by-doing condition they learned by performing writing tasks. Half of the students reflected on either the models' or their own performance. In our study, reflection did not affect academic writing performance and self-efficacy beliefs, and neither did instructional method. Both reflection and instructional method did influence students' satisfaction with the learning activities. Students preferred learning by doing over observational learning, and reflecting over not reflecting. From this study, we can conclude that in academic synthesis writing the interplay between reflection, observational learning and learning by doing is not evident yet: students seem to perform equally well in all conditions, even though they prefer learning by doing over observational learning, and reflecting over not reflecting.