The paper presents a Bayesian model of word learning that explains how people generalize meaning from a few examples of a novel word. The model is based on the idea that word learning involves Bayesian inference, where learners combine prior knowledge with statistical information from examples to infer the most likely meaning of a word. The model is tested against experimental data from adults and children, showing that it can account for the patterns of generalization observed in the data. The study involved 25 participants who were shown examples of words in a novel language and asked to identify other instances of those words. The results showed that participants generalized from one example to subordinate-level matches (e.g., other dalmatians) more often than to basic-level or superordinate-level matches. When given three examples, participants generalized more sharply, often only to the most specific category containing all three examples. The Bayesian model was applied to these data, and it was found that the model could accurately predict the patterns of generalization observed. The model incorporates a prior that favors more distinctive clusters and a likelihood that depends on the size of the cluster. The model also accounts for the fact that participants may have a basic-level bias, which is not captured in their similarity judgments. The paper also discusses the implications of these findings for theories of word learning. It suggests that a Bayesian framework can integrate both taxonomic and basic-level biases with statistical principles, providing a more comprehensive account of how people learn words. The study also highlights the importance of considering both the structure of the hypothesis space and the statistical properties of the examples in understanding word learning. The results are consistent with the idea that children have a basic-level bias in word learning, but this bias may develop later as children gain experience with how words are typically used. The study also suggests that the Bayesian model can be extended to studies of child learners and to the learning of words for novel objects. The findings have implications for understanding how people learn words and how this process is influenced by prior knowledge and statistical information from examples.The paper presents a Bayesian model of word learning that explains how people generalize meaning from a few examples of a novel word. The model is based on the idea that word learning involves Bayesian inference, where learners combine prior knowledge with statistical information from examples to infer the most likely meaning of a word. The model is tested against experimental data from adults and children, showing that it can account for the patterns of generalization observed in the data. The study involved 25 participants who were shown examples of words in a novel language and asked to identify other instances of those words. The results showed that participants generalized from one example to subordinate-level matches (e.g., other dalmatians) more often than to basic-level or superordinate-level matches. When given three examples, participants generalized more sharply, often only to the most specific category containing all three examples. The Bayesian model was applied to these data, and it was found that the model could accurately predict the patterns of generalization observed. The model incorporates a prior that favors more distinctive clusters and a likelihood that depends on the size of the cluster. The model also accounts for the fact that participants may have a basic-level bias, which is not captured in their similarity judgments. The paper also discusses the implications of these findings for theories of word learning. It suggests that a Bayesian framework can integrate both taxonomic and basic-level biases with statistical principles, providing a more comprehensive account of how people learn words. The study also highlights the importance of considering both the structure of the hypothesis space and the statistical properties of the examples in understanding word learning. The results are consistent with the idea that children have a basic-level bias in word learning, but this bias may develop later as children gain experience with how words are typically used. The study also suggests that the Bayesian model can be extended to studies of child learners and to the learning of words for novel objects. The findings have implications for understanding how people learn words and how this process is influenced by prior knowledge and statistical information from examples.