Eye Movements in Arabic Reading: Foveal and Parafoveal Processing

Measures of eye movements provide a moment-by-moment account of the visual and cognitive processes that underlie normal reading. These measures have been used to develop computational models that underlie reading for Latinate languages like English. However, little is known about the processes underlying reading for Semitic languages like Arabic that have different visual and linguistic characteristics. Thus, this thesis aims to examine the foveal and parafoveal processing of Arabic words during normal reading. The eight experiments herein provide a novel investigation of how linguistic, typographic, and visual features of Arabic influence eye movements during reading. Experiments 1 and 2 investigated effects of word predictability, finding that Arabic readers use contextual cues to facilitate word processing. The findings also indicated that readers engage in more detailed parafoveal processing when words are less morphologically complex, and when key letters are visible parafoveally. Experiments 3-5 investigated the effects of using monospaced fonts on word recognition and reading in Arabic. The results from these experiments indicated that Arabic readers experience greater difficulty when reading text in a monospaced font, and that this affected both lexical and contextual processing. Experiments 6-8 then aimed to investigate how letter shape information influences foveal and parafoveal processing in Arabic. The findings from Experiment 6 indicated that Arabic readers use the overall shape of words to identify misspelled words when these are available for foveal inspection. The findings from Experiments 7 and 8 indicated that Arabic readers extract partial letter information from the parafovea to facilitate reading. Overall, these studies highlight that language-specific features of Arabic play a key role in how Arabic readers move their eyes through text. These features include morphological complexity, word structure, inter-letter connectivity pattern, and letter shape similarity. Importantly, these findings will help in developing future computational models that could simulate reading in Arabic.