It is sometimes supposed that category learning involves competing explicit and procedural systems, with only the former reliant on working memory capacity (WMC). In 2 experiments participants were trained for 3 blocks on both filtering (often said to be learned explicitly) and condensation (often said to be learned procedurally) category structures. Both experiments (total N = 160) demonstrated that participants with higher WMC tended to be more accurate in condensation tasks, but not less accurate in filtering tasks. Furthermore, state-trace analysis did not find a differential influence of WMC on performance in these tasks. Finally, inspection of the mixture of response strategies at play across the 2 conditions and 3 blocks showed only a minor influence of WMC, and then only on later training blocks. The results provide no support for the existence of a "system" of category learning that is independent of working memory and are instead consistent with most single-system interpretations of category learning.