Distraction embedded in working memory tasks leads to impaired performance. This impairment is mitigated when targets and distractors that follow them share common features--a signature effect of interference by superposition. Here we propose that target-distractor similarity modulates not only forgetting from working memory but also encoding into long-term memory. In five experiments, we test this elaboration-by-superposition hypothesis, demonstrating that semantic relatedness between targets and distractors benefits delayed category-cued recall performance (Experiments 1a and 1b), which is not due to carry-over effects from working memory testing (Experiment 2). Just as in the case of working memory, this long-term memory effect is reduced when distractors precede targets (Experiment 3). Finally, we show that while high target-distractor similarity reduces forgetting from working memory, it produces net benefits for long-term memory performance (Experiment 4). Together, the results suggest that common mechanisms underlie encoding into working and long-term memory, and that bindings between features of spatiotemporal context and features of to-be-remembered items play a crucial role.