Eye Movements and Reading

When we read, we have the impression that our eyes glide smoothly across the page. However, this is an illusion. The eyes stay relatively fixed for about 200 to 300 milliseconds (ms) and then quickly jump (in about 20–30 ms) to a new location. The stable periods are called fixations and the eye movements are called saccades. Vision is suppressed during saccades (to prevent the perception of blur during eye movements) and useful information is obtained only during fixations. Our eyes typically move forward through the text, although regressions (saccades going backward through the text) account for about 10 to 15% of the movements. When readers come to the end of a line, a large leftward eye movement (called a return sweep) is made to near the beginning of the next line of text.

It is now easy to record eye movements accurately. The most precise eyetrackers have a spatial accuracy of less than one half a character, and they provide a millisecond-by-millisecond record in the time domain. Eyetrackers that are cheaper and easier to use are still spatially accurate to about a character and to about 5 ms temporally. Thus, with the types of tracker being used today, one can be quite sure which word a reader is fixating and can measure the duration of a fixation with an accuracy of less than 5 milliseconds (ms). This precision is important because what the eyes are fixating on is closely linked to the underlying cognitive processes of reading, as this entry describes.

The Perceptual Span in Reading

The most direct way that a tight linkage between the location of the eyes and the underlying processing in reading has been demonstrated is through the moving window paradigm first used in classic experiments by George McConkie and Keith Rayner. In this paradigm, what the reader can process from text on the video monitor depends on where the eyes are. For example, in a one-word window condition, the word that is being currently fixated is normal, but all the letters in the other words are replaced by Xs or random letters. However, when the eyes move and the reader fixates a new word, that word is now normal on the [Page 434]next fixation and all the other words (including the previously fixated word) have their letters replaced by Xs (see Figure 1a). Thus, this paradigm is like a custom “slide show,” where there is a new slide on each fixation (tailored to the fixation location). Provided that the window isn't less than a word, readers are generally not aware of the window; it just seems like normal reading.

Figure 1 The Moving Window and Boundary Paradigms

Notes: (a) An illustration of the moving window paradigm. The size of the window in the illustration is two words: the fixated word indicated by the asterisks and the word to the right. As indicated in the text, the size of the window can be defined either by the number of letters or the number of words. In addition, the location of the window can be manipulated so that, for example, a two-word window could include the fixated word and the word to the left. (b) An illustration of the boundary paradigm. The vertical mark indicates the boundary (which is not visible to the participant). The display change involves only a single word location, and the only display change occurs when the boundary is first crossed. In this illustration, a homophone of the target word is the preview. Other possible previews could be bench, broad, bsfhfs, or xxxxx. Fixation time on the target word is shorter when “beech” is the preview rather than “bench” (indicating phonological coding of the preview), but “bench” is a better preview than any of the others, indicating orthographic coding of the preview.

...