Non-foveal text processing and fixation positions in reading
This thesis reports seven experiments which investigate what determines where the eyes move during reading. Specifically, the experiments examine what kinds of linguistic information can influence where words are first fixated and refixated. Experiment 1 showed that fixations landed nearer to the beginning of words in which the initial letter sequence was misspelled, compared to when the words were spelled correctly. Experiments 2 and 3 showed that the effects of misspellings on saccade programming can not be explained by lexical non-foveal preprocessing, and therefore the results of Experiment 1 must be due to preprocessing of the orthographic familiarity of word initial letter sequences. These results were confirmed in Experiment 4, which showed that first fixations landed nearer to the beginning of correctly spelled words with orthographically irregular, compared to orthographically regular, initial letter sequences. Furthermore, Experiment 5 showed that these effects held for sentences presented in upper case text. Furthermore, Experiments 6 and 7 demonstrated that the influence of orthography on saccade programming was independent of foveal processing difficulty. These results are most consistent with an attraction based explanation in which preprocessing of orthography, independent of processing load, influences the word length and launch site based saccade programme to produce a small shift in the preferred viewing position in the direction of the orthographic irregularity. The results also show that linguistic processing can influence the direction and length of refixation saccades. Furthermore, although preprocessing of orthography can influence saccade programming, the results provide no consistent evidence for an influence of orthography on prior fixation durations or probabilities. These results indicate that there is independent processing of when and where the eyes move. The absence of robust "parafoveal-on-foveal" effects provides no support for parallel processing models of reading.