Research on active category learning—i.e., where the learner manipulates continuous feature dimensions of novel referents and receives labels for their self-generated exemplars—has routinely shown that people prefer to sample from regions of the stimulus space with high class uncertainty (near category boundaries). Prevailing accounts suggest that this strategy facilitates an understanding of the subtle distinctions between categories. However, prior work has focused on situations where category boundaries are rigid. In actuality, the boundaries between natural categories are often fuzzy or unclear. Here, we ask: do learners pursue uncertainty sampling when labels at the boundary are themselves uncertain? And how does the size of the label domain affect this?
To answer these questions, in two experiments we introduce a fuzzy buffer around a target category where conflicting labels are returned from two ‘teachers,’ and we evaluate how sampling and representation are affected. In experiment 1 we target a single-label domain while we employ a dual-label domain in experiment 2—despite maintaining an identically placed category boundary between experiments. Under the single-label domain (experiment 1), we find that fuzzy boundary learners avoid uncertainty, opting to sample densely from highly certain regions of the target category as opposed to its boundary, which held consequences for learners' grasp of the category structure even outside the fuzzy buffer zone. In experiment 2 we find that the availability of a second label resolves this preference for high-certainty sampling and alleviates the representational shortcomings associated with learning label meanings in the context of a fuzzy boundary. Our data show that multi-label contexts alter how learners negotiate learning in the presence of fuzzy class boundaries and suggest a powerful role for label contrast in the active development of word knowledge.