A fundamental task of language acquisition is to extract abstract algebraic
rules. Three experiments show that 7-month-old infants attend longer to se
ntences with unfamiliar structures than to sentences with familiar structur
es. The design of the artificial language task used in these experiments en
sured that this discrimination could not be performed by counting, by a sys
tem that is sensitive only to transitional probabilities, or by a popular c
lass of simple neural network models. Instead, these results suggest that i
nfants can represent, extract, and generalize abstract algebraic rules.