Judgments of learning (JOLs) are often reactive on memory for cue-target pairs. This pattern, however, is moderated by relatedness, as related but not unrelated pairs often show a memorial benefit compared to a no-JOL control group. Based on Soderstrom et al.'s, "Journal of Experimental Psychology: Learning, Memory, and Cognition" 41, 553-558, (2015) cue-strengthening account, JOLs direct attention towards intrinsic cues which aid retrieval. However, reactivity may also reflect specific processing of cue-target associations, which is applied whenever semantic associations are available, even when these associations are indirect. The present study tested this possibility using mediated associates (e.g., lion -- stripes) which are directly unrelated to each other and indirectly related through a non-presented mediator (e.g., tiger). Based on a cue-strengthening account, no reactivity would be expected for mediated associates. Alternatively, if cue strengthening primarily reflects enhanced processing of cue-target relations, memory benefits would be expected whenever pairs are semantically related, even if pairs are indirectly related through mediators. Overall, reactivity extended to mediated associates in cued-recall (Experiment 1) and recognition tests (Experiments 2 and 3). Interestingly, JOL reactivity was consistently found on recognition of non-mediated unrelated pairs (Experiments 2-4). Thus, positive reactivity on related pairs for cued-recall testing likely reflects increased activation of cue-target associations. However, because recognition is based on familiarity cues, reactivity occurs globally for all pair types, regardless of cue-target relations.