Mapping the Neurocognitive Architecture of Reading

Abstract

Reading relies on at least partially dissociable sub-word (sublexical) and word-specific (lexical) processes, which are subserved by somewhat separable neural substrates. Beyond this broad dissociation between sublexical and lexical reading, the specific knowledge and neural substrates that enable the reading of known and novel words are underspecified. This dissertation investigates the neurocognitive architecture of reading through psychometric and lesion-symptom mapping studies of reading aloud.

The first study used a psychometric approach to determine whether individual differences in adult reading skill reflect differences in knowledge of print-to-sound statistics. Literate adults read aloud a novel corpus of words varying in frequency, spelling-to-sound regularity, and imageability, as well as pseudowords varying in spelling-to-sound statistics. Mixed effects regressions examined group-effects and individual differences in reading aloud. Our results suggest that learned print-to-sound statistics determine both reading difficulty varying between words and reading skill varying between readers. Additionally, all three major components of word identity (spelling, pronunciation, and meaning) influence reading aloud.

The second study examined the neuroanatomical basis for acquired effects of lexicality, spelling-to-sound regularity, and concreteness on reading aloud in post-stroke aphasia. Support vector regression lesion-symptom mapping (LSM) revealed that phonological decoding of print relies on a left frontoparietal network that subserves sensory-motor integration. Within this frontoparietal network, accurate reading of pseudowords relies especially on left ventral precentral gyrus (lvPCG) being intact, suggesting that motor phonological processing is an essential prerequisite. Additionally, anterior inferior frontal gyrus may coordinate control over semantic contributions to reading aloud.

The final study examined whether impaired motor phonology preferentially disrupts reading aloud of pseudowords in post-stroke aphasia, as suggested by the second study. Mixed effects regression and a novel structural connectome LSM method confirmed an association between inaccurate pseudoword reading and both motor phonological impairment and disconnected lvPCG. In contrast, inaccurate reading of both pseudowords and words related to disconnected left temporoparietal cortex and impaired sensory-motor integration. These results motivate a neurocognitive model of reading that incorporates a sensory-motor phonological circuit.

Together, these three studies clarify the knowledge and neural substrates underlying reading and suggest new directions for psychological and neuroscientific studies of reading.