When building a lexicon, young children must first learn individual word-object mappings and subsequently extend these mappings to new category instances. Recent methodological advances demonstrate that processing efficiency during the initial mapping process increases between 15 and 24 months-of-age. The current study investigates real-time processing during the second word learning step, generalization. Using a head-mounted eye-tracker, 18-month old children completed a novel noun generalization task with novel solid objects. Many participants generalized names for solid objects based on similarity in shape rather than similarity in material. The addition of eye-tracking data in the current study reveals children’s comparison of test and exemplar objects both before and after novel name presentation. Integrating eye-tracking data with measures of vocabulary knowledge can elucidate how vocabulary organization speeds the generalization decision. Future work will manipulate syntactic context to examine how children’s knowledge of count and mass nouns further influences the generalization decision process.