It has been suggested that language learners prefer phonologically natural patterns over unnatural ones (McCarthy & Prince, 1995; Tesar & Smolensky, 1993, 2000). Various researchers have cited evidence for this assertion based on production (Demuth, 1995; Gnanadesikan, 1995) and perception (Jusczyk, Smolensky, & Allocco, 2002). Two experiments presented in this article test whether 9-month-old infants learn unmarked, phonetically grounded patterns more easily than marked, arbitrary ones. Experiment 1 tested phonetically grounded and arbitrary patterns involving manner of articulation; results revealed that infants showed no specific preference for the grounded patterns. Experiment 2 tested a pattern in place of articulation; once again, infants showed no specific preference for the grounded distributions. We take this as evidence that infants can, in fact, learn arbitrary phonological patterns and as evidence against a bias for grounded patterns in the grammar of infants at this age.