Learning to read has a substantial effect on the representations of spoken and meaning forms of words. In this paper we assess literacy effects beyond representational changes, focusing on adaptations to the architecture of the reading system that maps between these representations. We present a connectionist model of reading that predicted distinct processing of pre- and post-literacy acquired words. For reading for meaning, words learned prior to literacy were processed more indirectly via phonological representations, whereas for post-literacy acquired words, processing was more direct along the orthography to semantics pathway. This more computationally intensive route was prioritised because indirect phonology to semantics mappings were unavailable. Such an effect was less apparent for naming, because learning direct orthography to phonology mappings is less computationally intensive. These results were confirmed in an analysis of naming and lexical decision behavioural data. The effect of literacy onset remains an observable artefact in adult reading.