While such factors as demand characteristics, encoding, and retrieval inhibition were shown to be significant in producing the directed forgetting effect, no attention was paid to whether the intention to manage one’s own memory, per se, matters. In the present article, we addressed this important gap in the literature. To control the quality of encoding we ensured that both the to-be-remembered (TBR) and to-be-forgotten (TBF) items were genuinely learned before the manipulation. We used extremely long delays between the memory instructions and testing to release inhibition associated with the content of instructions. 98 participants demonstrated flawless recall of 12 Russian - made up language word pairs. They then viewed each Russian word from a pair once, with randomized instructions “Forget”, “Remember”, “Repeat”, or a short cognitive task. Self-reports on the mnemonic strategies were collected. Free recall and recognition tests were administered three times - 45 minutes, a month and a year (N = 58) later. Despite a strong incentive to recall all word pairs, fewer TBF pairs were recalled in comparison with TBR pairs, both after 45 minutes and after one month’s delay. Recognition among all conditions was equally high. A year later free recall was close to zero. In contrast, the TBR and TBF pairs were recognized equally better than pairs presented in “Repeat” and “Task” conditions. Thus, our results show that the intention to manage one’s own memory enhances the accessibility of memories at a very long time delay, no matter what type of instruction is issued.