<p>Humans have an exceptional ability to notice relations between different entities and situations and to transfer their relational knowledge across contexts. For example, adults can easily determine whether two objects are the same, regardless of what these objects are. It remains unclear which cognitive resources and strategies underlie human ability to acquire and use generalizable relational concepts. Here, in a set of three experiments (<i>N</i> = 40, <i>N</i> = 40, <i>N</i> = 155; data collected in 2021), we investigate the strategies that human adults use to determine whether two compound items are visually identical. We found that people compare the items by revealing perceptual information in the order that minimizes memory involvement and errors. When participants are prevented from using the perceptual offloading strategy, they switch to a naming strategy—either internally or through explicit verbalization—to maintain accuracy in relational judgment. When the items lack easily nameable features and perceptual offloading is restricted, participants are slower, less accurate, and less efficient in their comparisons. Thus, humans adaptively offload the memory requirements of the relational comparison onto active perceptual interactions with the stimuli and naming. We suggest that cognitive models of visual relational reasoning should consider these perceptual and language-based resources when formalizing mechanisms underlying human relational judgments.</p>