Working memory capacity (WMC) has been shown to correlate with performance on complex cognitive tasks, including language comprehension and production. However, some scholars have suggested that performance outcomes result from an interaction between individual differences (IDs), such as WMC, and learning conditions (Robinson, 2005a). Reber, Walkenfeld, and Hernstadt (1991) specifically claimed that IDs influence performance on explicit, but not implicit, processes. In this study, English native-speakers were exposed to a semi-artificial language under incidental or rule-search conditions, and their WMC was measured by two complex-span tasks. Both conditions produced a clear learning effect, with an advantage for the rule-search group. No significant correlations between overall performance on a grammaticality judgment task and WM scores were found for either group. However, WMC predicted performance on grammatical items for the rule-search group. These results support Reber et al.’s (1991) claim that aptitude measures may only be predictive of learning in explicit conditions.