The objective of this research is to identify ways in which computer simulations can be used to expedite the design of reading and spelling instruction. In many fields, computer models serve as surrogates for real-life situations. By running the models under various conditions, it is possible to evaluate theories and examine their utility, thus vastly increasing the likelihood that eventual real-life experimentation will be productive and cost-effective. Only recently have methods been devised for simulating some of the basic operations involved in learning to read. This research will extend these models, developed by Terence Sejnowski and James McClelland, to incorporate more of the natural complexities of reading than have been previously included. For example, the research will extend the models to study how the process of learning to read is affected by the way a word looks, the way it sounds, by the number of syllables, by simultaneous practice on spelling, and by first learning rules for pronouncing letter- sound units within words (phonics). The research will use these computer models to guide classroom experiments. For over 200 years there has been controversy over the ways in which such factors influence learning to read. As a result, "compromise" methods of teaching reading have been adopted and have proved ineffective with much of the population. Survey data have shown literacy in the United States to be at a dangerously low ebb, with fewer than 5% of our citizens reading well enough to handle college-level textbooks, for example. Knowledge gained by the use of advanced computer-simulation technology as a guide to experimentation can help to put the practice of reading instruction on a sound scientific footing.
