This paper presents a categorization experiment supporting the hypothesis that base-rate-neglect occurs in linguistic category learning (and, thus, is a cross-domain phenomenon), and that it is more likely for those learners who engage in explicit problem-solving, rather than implicit learning. We find that among those participants who were able to verbalize the cue that was probabilistically associated with category-membership (correct-staters) in a phonological learning task, about a third respond in a way consistent with base-rate-neglect. On the other hand, non-staters are more likely to respond randomly or by probability matching the base-rates. These results suggest that explicit learning is associated with base-rate-neglect to a greater extent. We found that no learners integrate the two probabilistic patterns present in the experiment into a single Bayesian estimate. Rather, some of them focus on the base-rates, others focus on category-internal cues, and others simply fail to learn anything.