ABSTRACT
Word recognition is primarily a bottom-up or data-driven process. The 'top-down' knowledge of the word is affecting the perception of the speech stream. Recognising printed words presents a different set of problems from understanding spoken words. This chapter examines briefly a couple of models of speech recognition. The cohort model of William Marslen-Wilson emphasises the way speech unfolds across time. The modellers make strong claims that taken together make the triangle model a radical alternative to traditional models of reading and word recognition. The triangle model does a good job of explaining normal reading, and explains surface dyslexia in terms of disruption to the semantic-phonology pathway. The full alphabetic phase children have full knowledge of letters and sounds and how they correspond, so they can read words they've never seen before. The different senses of most ambiguous words aren't equally commonly intended.
