An Investigation of Reading Development Through Sensitivity to Sublexical Units

The present dissertation provides a novel perspective to the study of reading, focusing on sensitivity to sublexical units across reading development. Work towards this thesis has been conducted at SISSA and Macquarie University. The first study is an eye tracking experiment on natural reading, with 140 developing readers and 33 adult participants, who silently read multiline passages from story books in Italian. A developmental database of eye tracking during natural reading was created, filling a gap in the literature. We replicated well-documented developmental trends of reading behavior (e.g., reading rate and skipping rate increasing with age) and effects of word length and frequency on eye tracking measures. The second study, in collaboration with Dr Jon Carr, is a methodological paper presenting algorithms for accuracy enhancement of eye tracking recordings in multiline reading. Using the above-mentioned dataset and computational simulations, we assessed the performance of several algorithms (including two novel methods that we proposed) on the correction of vertical drift, the progressive displacement of fixation registrations on the vertical axis over time. We provided guidance for eye tracking researchers in the application of these methods, and one of the novel algorithms (based on Dynamic Time Warping) proved particularly promising in realigning fixations, especially in child recordings. This manuscript has recently been accepted for publication in Behavior Research Methods. In the third study, I examined sensitivity to statistical regularities in letter co-occurrence throughout reading development, by analysing the effects of n-gram frequency metrics on eye-tracking measures. To this end, the EyeReadIt eye-tracking corpus (presented in the first study) was used. Our results suggest that n-gram frequency effects (in particular related to maximum/average frequency metrics) are present even in developing readers, suggesting that sensitivity to sublexical orthographic regularities in reading is present as soon as the developing reading system can pick it up – in the case of this study, as early as in third grade. The results bear relevant implications for extant theories of learning to read, which largely overlook the contribution of statistical learning to reading acquisition. The fourth study is a magnetoencephalography experiment conducted at Macquarie University, in collaboration with Dr Lisi Beyersmann, Prof Paul Sowman, and Prof Anne Castles, on 28 adults and 17 children (5th and 6th grade). We investigated selective neural responses to morphemes at different stages of reading development, using Fast Periodic Visual Stimulation (FPVS) combined with an oddball design. Participants were presented with rapid sequences (6 Hz) of pseudoword combinations of stem/nonstem and suffix/nonsuffix components. Interleaved in this stream, oddball stimuli appeared periodically every 5 items (1.2 Hz) and were specifically designed to examine stem or suffix detection (e.g., stem+suffix oddballs, such as softity, were embedded in a sequence of nonstem+suffix base items, such as terpity). We predicted that neural responses at the oddball stimulation frequency (1.2 Hz) would reflect the detection of morphemes in the oddball stimuli. Sensor-level analysis revealed a selective response in a left occipito-temporal region of interest when the oddball stimuli were fully decomposable pseudowords. This response emerged for adults and children alike, showing that automatic morpheme identification occurs at relatively early stages of reading development, in line with major accounts of morphological decomposition. Critically, these findings also suggest that morpheme identification is modulated by the context in which the morphemes appear.