Think/No-Think studies have shown that people can prevent memories from coming up to their mind by explicitly attempting to not think of them. However, there is an important limitation in the measures typically used: binary recall vs. no-recall accuracy under a specific time deadline. In this study, we instead focused on recall latency with a longer response window to accommodate a wider range of recall latencies. We found in Experiment 1 that direct suppression in the standard No-Think condition had a relatively uniform, graded effect, slowing the recall process in such a way that more recall failures occur with a short deadline, but a longer deadline (10 s) allows for successful recall at rates comparable to a baseline condition. In Experiment 2, thought substitution also caused the slowdown of the recall despite still lower recall rate than a baseline in 10 s. These results suggest that memory recall is subject to graded impairment across all items in a consistent manner instead of the breakdown of recall among a subset of memories. For understanding forgetting by retrieval stopping, excessive use of recall rate should be avoided, and recall latency is a potential alternative.