We develop and test a dual-route model of genetic effects on reading aloud and spelling, based on irregular and non-word reading and spelling performance assessed in 1382 monozygotic and dizygotic twins. As in earlier research, most of the variance in reading was due to genetic effects. However, there were three more specific conclusions: the first was that most of the genetic effect is common to both regular and irregular reading. In addition to this common variance evidence was found for distinct genes influencing the acquisition of a lexicon of stored words, and additional genetic effects influencing the acquisition of grapheme–phoneme correspondence rules. The third conclusion, from a combined model of reading and spelling, is that reading and spelling have a common genetic basis. Models that did not distinguish lexical and non-lexical performance fit significantly worse than dual route genetic models. An implication of the research is that models of reading, whether connectionist or dual-route, must allow for the genetic independence of neurological processes underlying the decoding of non-words and irregular words.