There is accumulating evidence that making judgments of learning (JOLs) while studying word pairs improves memory for related but not for unrelated pairs. To explain these findings, it has been proposed that people process cue-target relatedness more when making JOLs than they do spontaneously. We directly tested this relatedness-processing assumption with unrelated and related word pairs as well as with identical word pairs, a type of material that has not yet been used in JOL reactivity research. In three experiments, participants did or did not make JOLs while studying word pairs for a cued-recall test. They also indicated at test whether each cue appeared at study with an unrelated, a related, or an identical target before attempting to recall it (relatedness judgments). Results revealed that making JOLs improved memory for related and identical word pairs but not for unrelated word pairs. Independently from these effects on memory performance, making JOLs improved the accuracy of relatedness judgments. These findings provide direct evidence for the relatedness-processing assumption and thereby extend our understanding of JOL reactivity. More generally, this study suggests that instructions to monitor one’s learning can direct attention to information that is not or less processed otherwise.