For decades, theories of early word learning have assumed that infants are equipped with learning biases that help them learn words at a fast pace. One of these biases, called Mutual Exclusivity, suggests that infants reject second labels for name-known objects. Our first two experiments, with children and with infants, suggest that novelty preference during Mutual Exclusivity tasks should not be taken as evidence that associations between novel labels and name-known objects have not taken place. A third experiment, supplemented with computational modeling, ruled out cascaded activation patterns as alternative explanations and, instead, confirmed that word-object associations are non-selective throughout infancy and childhood.