The relevance of Broca’s area to both language and non-linguistic sequence processing is well established. However, manyof the previous fMRI studies on artificial grammar learning use letter sequences as their non-linguistic stimuli. Since theletters are linguistic in nature, these may inadvertently activate language circuits independent of the artificial grammar. Inaddition, participants have been explicitly told before testing that they needed to classify sentences as either grammaticalor ungrammatical. Thus, it is possible that part or all of the activation of Broca’s reported in these studies is an artifact ofthese manipulations. In our current study, we used sequences of human faces instead of letters, and tested participants insuch a way that they were never aware they were even being tested. Nevertheless, most participants still showed evidenceof learning the non-linguistic artificial grammar, and their Broca’s area was also differentially active for ’grammatical’ vs.’ungrammatical’ sequences.