However, just as 'bottom up' phonic models have been challenged, so too are the 'top down' models. As 'meaning' is the central premise, the clause or sentence is seen as 'more significant linguistically than the word or letter'. (Wray & Medwell, p.100). The reader would not have to foster any actual knowledge of the features of written text. The main problem here is that '...research has demonstrated that the ability to segment words into phonemes may predict later reading ability'. (Wray & Medwell, p.100) This would suggest that a knowledge and awareness of written language features (and hence of phonics) must have a vital role to play in the process of learning to read.
Interactive models have now been suggested which attempt to recognise that neither 'top down' nor 'bottom up' theories can represent the whole picture and which attempt to use the strengths of both. A model such as this was proposed by Rummelhart (1977) who saw reading as a synthesis of the two approaches. This new type of model allows both previous experience and 'code' features of written text to work together for meaning to be created. …