Languages and writing systems result from satisfying multiple constraints related to learning, comprehension, production, and their biological bases. Orthographies are not optimal because these constraints often conflict, with further deviations due to accidents of history and geography. Things tend to even out because writing systems and the languages they represent exhibit systematic trade-offs between orthographic depth and morphological complexity.

We distinguish between literal and metaphorical applications of Bayesian models. When intended literally, an isomorphism exists between the elements of representation assumed by the rational analysis and the mechanism that implements the computation. Thus, observation of the implementation can externally validate assumptions underlying the rational analysis. In other applications, no such isomorphism exists, so it is not clear how the assumptions that allow a Bayesian model to fit data can be independently validated.

Reading educators and clinicians dating from Orton (1928) have asserted that dyslexic children exhibit impairments in one or more aspects of the reading process that are not seen in nondyslexic children. There has been chronic disagreement about the identity and causes of these impairments, but their existence has been generally assumed. We will term this the standard view of dyslexia.

Children assessed as reading disabled are often thought to use decoding processes that differ from those of nondisabled children. This assumption was examined in a study that compared the word recognition skills of a group of clinic-diagnosed reading disabled children with those of good and poor readers. Subjects read words and nonwords containing either regular or homographic spelling patterns. Regular patterns have a single pronunciation (e.g., -UST) while homographic patterns have multiple pronunciations (e.g., -ONE). Analyses of the errors, latencies, and types of pronunciations indicated that while the performance of the poor and disabled readers differed from that of the good readers, the two below-average reader groups were very similar. The reading disabled children exhibited decoding processes similar to those exhibited by younger nondisabled readers. The results suggest that many children who meet the diagnostic criteria for reading disability may be indistinguishable from nondisabled children in terms of actual reading performance.