The present study explores morphological bootstrapping in cross-situational word learning. Adult, English-speaking participants were exposed to novel words from an artificial language from three different semantic categories: fruit, animals, and vehicles. In the Experimental conditions, the final CV syllable was consistent across categories (e.g., /-ke/ for fruits), while in the Control condition, the endings were the same, but were assigned to words randomly. After initial training on the morphology under various degrees of referential uncertainty, participants were given a cross-situational word learning task with high referential uncertainty. With poor statistical cues to learn the words across trials, participants were forced to rely on the morphological cues to word meaning. In Experiments 1-3, participants in the Experimental conditions repeatedly outperformed participants in the Control conditions. In Experiment 4, when referential uncertainty was high in both parts of the experiment, there was no evidence of learning or making use of the morphological cues. These results suggest that learners apply morphological cues to word meaning only once they are reliably available.