Modeling Developmental Word Learning in Late Talking Children

In typical early vocabulary development, young toddlers are skilled at learning noun categories. In fact, they use attentional word learning biases in order to categorize nouns, which helps increase their vocabulary. In particular, in a Novel Noun Generalization task, they show a shape bias for solid objects, and a material bias for non-solid substances. Children who are at a delayed vocabulary level compared to their peers, or late talkers, do not exhibit the same word learning biases. A sample of 33 late talkers showed that those who remain persistently delayed in the long term differ from other late talkers in both their noun vocabulary structures and their word learning biases. Behavioral data using the Novel Noun Generalization task shows that the group of long term persistent late talkers developed a shape bias that was not significantly larger than the material bias, whereas the other two groups of late talkers had a significantly larger shape bias than material bias. Computational models that simulate the Novel Noun Generalization task show that the long term persistent late talkers develop a shape bias and a material bias faster than the other two groups. This tell us that this particular group of late talkers stands out from the others, and may not develop biases in the same way, which could be the cause of their vocabulary delay. The direction of this difference is not entirely clear, as the models do not replicate the behavioral data. However, acknowledging a fundamental difference in the way different types of late talkers learn new words is a step toward creating early interventions, and further replications of these studies could help the network simulations further match behavioral data seen in the lab. Word Learning in Late Talking Children 3