Beginning with Fitch and Hauser (2004), a number of studies have used the Artificial Grammar Learning task to investigate learning rules generating hierarchical structural relations among sequences of elements that are characteristic of the grammar of human languages. Studies that have examined the learning of a center-embedding rule (AnBn rule) exemplified by the sentence, The dogs the girl the boys like feeds bark incessantly have provided mixed results. We present the results of three experiments that demonstrate learning when training occurs incrementally (e.g., Lai & Poletiek, 2011) and requires feedback when testing with a grammaticality judgment task. We also use a novel completion task, which demonstrated learning both with and without feedback. In all cases, not all participants learned the rule.