In recent years, there has been considerable interest in how to maximize learners' retention of multiword expressions. One technique that has been shown to be highly effective is the use of exercises such as those found in mainstream English as a second language textbooks. In the present study, we investigated how the execution of a gap-fill exercise impacts the learning of phrasal verbs with 118 learners studying English as a foreign language. Participants completed a gap-fill exercise by referring to the answer key, or they received the answer key only after completing the exercise. The effects of the learning conditions were assessed with tests for measuring productive and receptive knowledge at two retention intervals. The results from mixed-effects logistic regression modeling showed that both executions of the gap-fill exercise led to similar rates of retention. The findings largely challenge previous research. We also explored how to minimize proactive interference when participants make errors in gap-fill exercises by asking them to recollect their initial guesses during the posttests. The results showed that when the initial guess was produced, correct recall of the target phrasal verbs was much greater than when the guess was not recollected. The finding indicates that memory for the initial guess may play a vital role in how participants learn from their errors. The pedagogical implications of the findings are discussed, and future areas of research are proposed.