How does Harm & Seidenberg's account of developmental dyslexia correspond to/refine the view presented in Snowling (1991)? What sort of insights do you get (if any) from this sort of computational account.