This component describes an experiment aimed at testing the CR variability effect when words/pairs are highly- and meaningfully-related. We hoped to a) examine the effects of relatedness, and b) potentially create a wordset with higher CR performance that we could use to address the floor effects in Experiments 2a and 2b. **Using highly-related word pairs from the DRM norms** We created a new set of words using DRM norms, where pairs consisted of probes (as cues) and lures (as CR and FR targets). Cues for each lure were the probe words with the 3rd highest probe-lure backwards associative strength. We selected a pool of 31 word pairs where the 3rd highest backwards associative strength exceeded .10. For an initial pilot, we tested *N* = 20 Prolific participants on *either* CR or FR, giving each participant a single study-test cycle of 15 words (pairs) randomly selected from the wordset. Exclusions were much lower--only 1 participant excluded for a major distraction and 1 reporting understanding less than 75% of the words. Of note, *no* participants were excluded for failing to get at least one right on their assigned list. As for accuracy (and variability): @[osf](dr7jq) FR performance still allows for a decent range, but CR performance is about as ceiling as it can get. Perhaps we can try with 4th, 5th-strongest probes, etc. **Using weaker DRM associates** We opted to use new, weaker probes (cues), specifically, those with the 10th-strongest (out of 15) BAS. For simplicity and efficiency, participants only received a single CR study-test cycle of 15 word pairs. We tested *N* = 29 participants and after excluding 8 for technical issues (a fixed programming error), 1 who reported a major distraction, and 2 who reported completing a similar study on another platform, we ended up with *N* = 18 participants. Their performance: @[osf](wcuxs) Better, but still higher than FR performance in all our other experiments. We might consider moving to the extreme end of the list (e.g., weakest associates?). Worth noting that 15 out of 42 of the current pairs already have the minimum BAS (i.e., 0), so it's not clear that changing the remaining 27 pairs would drastically affect performance.