This study aimed to examine how variable exposure to language statistical patterns affects reading behaviour, specifically, eye-movements during reading. The statistical patterns of language affect how individuals store, produce and comprehend language. When reading, individuals with greater linguistic proficiency typically have been shown to rely less on language statistical information compared to less proficient readers. Based on the Lexical Quality Hypothesis, however, it was hypothesized that spelling bias, a print-specific probabilistic cue, may only be utilized for representations with sufficient strengths of representation - through increased exposure to print in individuals, or through higher frequency of occurrence for individual words, since these individuals, and these words, would be expected to have representations of high quality in the reader’s mental lexicon. Undergraduate students with varying amounts of reading experience were presented with sentences containing English noun-noun compound words that varied in spelling bias, i.e. the probability of occurring in text either as spaced (window sill) or concatenated (windowsill). Linear mixed effect multiple regression models were fitted to the eye-movement data and demonstrated that compound words presented in their more supported format - i.e. the format with the highest bias, were read faster, but that this effect was modulated by reading experience, as measured by a test of exposure to print, as well as by word frequency. Only individuals with the most reading experience, and words with the highest frequencies benefited from this facilitatory effect of bias. This distributional property can thus be used during reading, but only when individuals' lexical representations are of sufficiently high quality. The results of this study thus suggest that future research considering the relationship between linguistic properties and reading must consider individual differences in reading skill and exposure.

Master of Science (MSc)

Thesis