| Summary: | This thesis investigates the strategies infants use to generalise labels to different objects in the early stages of lexical development. It aims to directly test the assumption that a taxonomic bias exists which guides infants to extend words to categories of objects instead of individual instances of them, against the hypothesis that infants discover the extension of words through exposure to multiple naming instances of different objects.Experiments One and Two attempted to teach two object-label pairings to infants at the end of their first year of life, and test generalisation of those labels to new objects from the same adult linguistic categories. This aim failed because infants showed evidence for prior knowledge of the words. Experiments Three and Four employed a more infant-controlled procedure using a habituation task during training; in the former a single exemplar from each adult category was used, whereas in the latter multiple exemplars from each category were used. In both Experiments evidence for word learning was provided at test, but infants failed to generalise the labels to other objects. Experiment Five used a training phase identical to Experiment Four but tested infants for perceptual categorisation in the absence of any labels. Some infants showed evidence for their ability to create such categories on the basis of the training set, suggesting that the inability to generalise in Experiments Three and Four was not due to a perceptual limitation. These findings suggest that infants at the end of their first year do not seem to be guided by any linguistic biases in their generalisation of labels. This thesis concludes that 10-month-old infants seem to have more advanced linguistic abilities than has previously been thought and constraint-like behaviour in later stages of lexical development might be a result of experience instead of a qualitative shift in cognitive processes.
|
|---|
