For decades, theories of early word learning have assumedthat infants are equipped with learning biases that help themlearn words at a fast pace. One of these biases, called MutualExclusivity, suggests that infants reject second labels forname-known objects. Our first two experiments, with childrenand with infants, suggest that novelty preference duringMutual Exclusivity tasks should not be taken as evidence thatassociations between novel labels and name-known objectshave not taken place. A third experiment, supplemented withcomputational modeling, ruled out cascaded activationpatterns as alternative explanations and, instead, confirmedthat word-object associations are non-selective throughoutinfancy and childhood.