Abstract

Two experiments explored learning and generalization in a novel orthography. In these experiments, adults were exposed to the written and spoken forms of novel words written in novel symbols. Novel words varied in the consistency and frequency of their symbol-sound mappings. Learners were not taught about individual symbol-sound mappings only whole word pronunciations. Experiment 1 demonstrated that learners can extract context-dependent symbol-sound mappings through exposure to whole novel words. They were also able to use this knowledge to read a further set of novel words in a generalization task. The frequency and consistency of symbol-sound mappings influenced learning and generalization in ways that largely mirrored effects seen in English orthography. In Experiment 2 learners were pre-exposed to novel word sounds. For half the items they also learned a meaning in the form of a novel definition. In the initial stages of orthographic learning, pre-exposure to either word sounds or word meanings provided equivalent benefit. However, by the end of training, a significant advantage for the meanings condition emerged. These experiments show that an artificial orthography paradigm can be successfully used to investigate the process of learning to read. Experiment 2 supported the idea that semantic knowledge influences learning to read aloud but also highlighted the importance of familiarity with whole-word phonological form. In my current work I am combining artificial orthography learning with neuroimaging to examine the neural substrates of learning to read.