This article examines the role of subjective familiarity in the implicit and explicit learning of artificial grammars. Experiment 1 found that objective measures of similarity (including fragment frequency and repetition structure) predicted ratings of familiarity, that familiarity ratings predicted grammaticality judgments, and that the extremity of familiarity ratings predicted confidence. Familiarity was further shown to predict judgments in the absence of confidence, hence contributing to above-chance guessing. Experiment 2 found that confidence developed as participants refined their knowledge of the distribution of familiarity and that differences in familiarity could be exploited prior to confidence developing. Experiment 3 found that familiarity was consciously exploited to make grammaticality judgments including those made without confidence and that familiarity could in some instances influence participants' grammaticality judgments apparently without their awareness. All 3 experiments found that knowledge distinct from familiarity was derived only under deliberate learning conditions. The results provide decisive evidence that familiarity is the essential source of knowledge in artificial grammar learning while also supporting a dual-process model of implicit and explicit learning. (Contains 2 figures, 18 tables, and 5 footnotes.)