The present review summarizes research investigating how words are identified parafoveally (and foveally) in reading. Parafoveal and foveal processing are compared when no other concurrent task is required (e.g., in singleword recognition tasks) and when both are required simultaneously (e.g., during reading). We first review methodologies used to study parafoveal processing (e.g., corpus analyses and experimental manipulations, including gaze-contingent display change experiments such as the boundary, moving window, moving mask, and fast priming paradigms). We then turn to a discussion of the levels of representation at which words are processed (e.g., orthographic, phonological, morphological, lexical, syntactic, and semantic). Next, we review relevant research regarding parafoveal processing, summarizing the extent to which words are processed at each of those levels of representation. We then review some of the most controversial aspects of parafoveal processing, as they relate to reading: (1) word skipping, (2) parafoveal-on-foveal effects, and (3) n + 1 and n + 2 preview benefit effects. Finally, we summarize two of the most advanced models of eye movements during reading and how they address foveal and parafoveal processing.

One of the words that readers of English skip most often is the definite article the. Most accounts of reading assume that in order for a reader to skip a word, it must have received some lexical processing. The definite article is skipped so regularly, however, that the oculomotor system might have learned to skip the letter string t-h-e automatically. We tested whether skipping of articles in English is sensitive to context information or whether it is truly automatic in the sense that any occurrence of the letter string the will trigger a skip. This was done using the gaze-contingent boundary paradigm (Rayner, 1975) to provide readers with false parafoveal previews of the article the. All experimental sentences contained a short target verb, the preview of which could be correct (i.e., identical to the actual subsequent word in the sentence; e.g., ace), a nonword (tda), or an infelicitous article preview (the). Our results indicated that readers tended to skip the infelicitous the previews frequently, suggesting that, in many cases, they seemed to be unable to detect the syntactic anomaly in the preview and based their skipping decision solely on the orthographic properties of the article. However, there was some evidence that readers sometimes detected the anomaly, as they also showed increased skipping of the pretarget word in the the preview condition.

In the boundary change paradigm (Rayner, 1975), when a reader's eyes cross an invisible boundary location, a preview word is replaced by a target word. Readers are generally unaware of such changes due to saccadic suppression. However, some readers detect changes on a few trials and a small percentage of them detect many changes. Two experiments are reported in which we combined eye movement data with signal detection analyses to investigate display change detection. On each trial, readers had to indicate if they saw a display change in addition to reading for meaning. On half the trials the display change occurred during the saccade (immediate condition); on the other half, it was slowed by 15-25 ms (delay condition) to increase the likelihood that a change would be detected. Sentences were presented in an alternating case fashion allowing us to investigate the influence of both letter identity and case. In the immediate condition, change detection was higher when letters changed than when case changed corroborating findings that word processing utilizes abstract (case independent) letter identities. However, in the delay condition (where d' was much higher than the immediate condition), detection was equal for letter and case changes. The results of both experiments indicate that sensitivity to display changes was related to how close the eyes were to the invalid preview on the fixation prior to the display change, as well as the timing of the completion of this change relative to the start of the post-change fixation.

The use of gaze-contingent display techniques to study reading has shown that readers attend not only the currently fixated word, but also the word to the right of the current fixation. However, a critical look at the literature shows that a number of questions cannot be readily answered from the available literature reviews on the topic. First, there is no consensus as to whether readers also attend the second word to the right of fixation. Second, it is not clear whether parafoveal processing is more efficient in languages such as Chinese. Third, it is not well understood whether the measured effects are confounded by the properties of the parafoveal mask. In the present study, we addressed these issues by performing a Bayesian metaanalysis of 93 experiments that used the boundary paradigm (Rayner, Cognitive Psychology, 7, 65-81. doi:10.1016/0010-028590005-5, 1975. We describe three main findings: (1) The advantage of previewing the second word to the right is modest in size and likely is not centered on zero; (2) Chinese readers do seem to make more efficient use of parafoveal processing, but this is mostly evident in gaze durations; and (3) there are interference effects associated with using different parafoveal masks that roughly increase when the mask is less word-like.Keywords Parafoveal processing . Reading . Preview benefit . Perceptual span . Eye movementsThe advance of eye-tracking technology has allowed an unprecedented opportunity to understand how the reading process unfolds in space and time. One of the advantages of this method is the possibility to precisely manipulate what participants see in real time while they are reading sentences on the screen. The use of such techniques has shown that readers not only process the currently fixated word, but that they also benefit from previewing the upcoming word in parafoveal vision (Rayner, 1998). This so-called preview benefit effect is measured as shorter fixation durations on a word for which correct preview information was available during the preceding fixation.The preview benefit effect is arguably one of the most robust and least controversial findings in the literature, and it has inspired many subsequent experiments as well as models of reading (Schotter, Angele, & Rayner, 2012). A particular issue highlighted by the preview benefit effect is the distinction between gaze location and the attentional focus during reading. Although the gaze location and attentional focus are usually identical in single-word recognition tasks, readers appear to routinely attend the upcoming word as well. Many experiments have been dedicated to determining exactly which properties of an upcoming word readers can process while they are still fixating on the preceding word. It has been shown that reader can preprocess the orthographic, phonological, morphological, and possibly semantic properties of the upcoming word (for a review, see Schotter et al., 2012).However, despite four decades of research on the topic, the size of this effect has not been system...

In research on eye movements in reading, it is common to analyze a number of canonical dependent measures to study how the effects of a manipulation unfold over time. Although this gives rise to the well-known multiple comparisons problem, i.e. an inflated probability that the null hypothesis is incorrectly rejected (Type I error), it is accepted standard practice not to apply any correction procedures. Instead, there appears to be a widespread belief that corrections are not necessary because the increase in false positives is too small to matter. To our knowledge, no formal argument has ever been presented to justify this assumption. Here, we report a computational investigation of this issue using Monte Carlo simulations. Our results show that, contrary to conventional wisdom, false positives are increased to unacceptable levels when no corrections are applied. Our simulations also show that counter-measures like the Bonferroni correction keep false positives in check while reducing statistical power only moderately. Hence, there is little reason why such corrections should not be made a standard requirement. Further, we discuss three statistical illusions that can arise when statistical power is low, and we show how power can be improved to prevent these illusions. In sum, our work renders a detailed picture of the various types of statistical errors than can occur in studies of reading behavior and we provide concrete guidance about how these errors can be avoided.