Eye Movements During Reading Research Paper

1. Basic Facts About Eye Movements During Reading

Although it seems as though we move our eyes smoothly across the text when we read, this is an illusion since our eyes make a series of movements (called saccades) that last about 20–35 ms. Between the saccades, our eyes are relatively still for about 200–250 ms (these periods when the eyes are stationary are called fixations). No information is obtained during saccades because the eyes are moving so quickly across the text; all new information is thus encoded during the fixations. Reading is therefore a bit like a slide show where each new slide (with a different view of the text) appears every quarter of a second or so. The saccade is typically 7–9 letter spaces and is not affected by the size of the print as long as it is not too small or too large. The appropriate metric to use when discussing eye movements therefore is letter spaces, and not visual angle (generally, 3–4 letter spaces are equivalent to 1 degree of visual angle). The other primary characteristic of eye movements is that about 10–15 percent of the time readers move their eyes back in the text (regressions) to look at material that has already been read.

Eye movements during reading are necessary because of acuity limitations in the visual system. With respect to the fixation point, any line of text can be divided into three regions: foveal, parafoveal, and peripheral. In the foveal region (extending 1 degree of visual angle to the left and right of fixation), acuity is sharpest and the letters can be easily resolved. In the parafoveal region (extending to 5 degrees of visual angle on either side of fixation) and the peripheral region (everything on the line beyond the parafoveal region), acuity drops off markedly so that our ability to identify letters is not very good even in the near parafovea. The purpose of eye movements in reading therefore is to place the fovea on that part of the text to be processed next. Whereas a majority of the words in a text are fixated during skilled reading, many words are skipped so that foveal processing of each word is not necessary. For example, content words are fixated about 85 percent of the time while function words are fixated about 35 percent of the time. Function words are fixated less frequently than content words because they are generally quite short and there is a relationship between word skipping and word length: as length increases, the probability of fixating increases; whereas 2–3 letter words are fixated less than 25 percent of the time, words that are 8 letters or longer are almost always fixated (and often fixated more than once).

When reading passages of text that are hard or unfamiliar, the eye movements of skilled readers are somewhat different from when they read text that is easy or familiar. As text difficulty increases, there are systematic effects on eye movements: fixation durations increase, saccade lengths decrease, and regression frequency increases. An important fact about eye movements in reading is that the ease or difficulty associated with understanding the text strongly influences the durations of fixations, the length of saccades, and the frequency of regressions.

It is important to note that the values presented above for fixation duration, saccade length, and regression frequency are averages and there is considerable variability in all of the measures. Thus, although the mean fixation duration is 200–250 ms and the mean saccade length is about 8 letter spaces, for individual readers these values might be somewhat higher or lower. Thus, the mean fixation duration for individual readers ranges between 200 ms and 300 ms and the average saccade length ranges from 6 to 10 letters. This between-reader variability (which also exists for regression frequency) is perhaps not as important as the fact that there is considerable withinreader variability. In other words, although a reader’s mean fixation duration is 250 ms, the range of individual fixation durations can be from under 100 ms to over 500 ms within a passage of text. Likewise, the variability in saccade length can range from 1 letter to over 15 letter spaces (though such long saccades typically follow regressions).

While fixation durations are typically examined when global aspects of reading are considered, when more detailed analyses involving individual target words are necessary, measures other than the mean fixation duration are usually examined. If readers always made one and only one fixation on a word then the fixation duration on the word would be the unit of processing. However, words are sometimes fixated more than once, and sometimes they are skipped. Thus, using the mean fixation duration to infer the processing time for a word is inadequate because it underestimates the time the eyes are actually on the word (that is, a 200 ms fixation and a 150 ms fixation on the same word would yield a mean fixation duration of 175 ms). Likewise, using only words that are fixated once (single fixation duration) is problematic because many words are skipped. Thus, the two most frequently used measures of processing time for a word are first fixation duration (the duration of the first fixation on a word independent of the total number of fixations) and gaze duration (the duration of all fixations on a word prior to an eye movement to another word).

2. Developmental Changes In Eye Movements

When children learn to read, they must learn to systemically move their eyes through text. Children in first grade typically fixate for 300–400 ms and only move their eyes a few characters. Thus, they fixate on virtually every word in the text. However, as reading skill develops, eye movements become more like those of skilled readers. By the time children are in fourth or fifth grade, the basic characteristics of their eye movements are like those of skilled readers. While their fixation durations and saccade lengths are similar to those of skilled readers, they do make more regressions than skilled readers. Children who experience severe difficulties learning to read (dyslexic readers) do not follow this systematic progression in the development of eye movement characteristics. Indeed, adult dyslexic readers make longer fixations, shorter saccades, and more regressions than skilled readers.

3. The Perceptual Span

How much information does a skilled reader acquire on each fixation and what is the size of the perceptual span (or area of effective vision) on each fixation? In order to investigate this question, the eye-contingent display change paradigm (McConkie and Rayner 1975, Rayner 1975, Rayner and Bertera 1979) was developed. In this paradigm, a reader’s eye movements are monitored (generally every millisecond) by a highly accurate eye-tracking system. The eyetracker is interfaced with a computer which controls the display monitor from which the reader reads and changes in the text are made contingent on the location of the reader’s eyes. Generally, the display changes are made during saccades and the reader is not aware of the changes.

There are three primary types of eye-contingent paradigms: the moving window, foveal mask, and boundary techniques. With the moving window technique, on each fixation a portion of the text around the fixation point is available to the reader. However, outside of this window, the text is replaced by other letters, or by Xs (see Fig. 1). When the reader moves his or her eyes, the window moves with the eyes. Thus, wherever the reader looks, there is readable text within the window and altered text outside the window. The rationale with the technique is that when the window is as large as the region from which a reader can normally obtain information, reading will not differ from when there is no window present. The foveal mask technique is very similar to the moving window paradigm except that the text and replaced letters are reversed. Thus, wherever the reader looks, the letters around the fixation are replaced by Xs while outside of the mask area the text remains normal (see Fig. 1). Finally, in the boundary technique, an invisible boundary location is specified in the text and when the reader’s eye movement crosses the boundary, an originally displayed word or letter string is replaced by a target word (see Fig. 1). The amount of time that the reader looks at the target word is computed as a function of (a) the relationship between the initially displayed stimulus and the target word, and (b) the distance that the reader was from the target word prior to launching a saccade that crossed the boundary location.

For readers of English (and other alphabetic writing systems printed from left-to-right), the span extends from the beginning of the currently fixated word, or roughly 3–4 letter spaces to the left of fixation, to about 14–15 letter spaces to the right of fixation. The span is thus asymmetric—it extends further to the right of fixation than to the left of fixation. For readers of languages printed from right-to-left (such as Hebrew), the span is asymmetric but in the opposite direction from English so that it is larger left of fixation than right. While the span is asymmetric, it is the case that no useful information is acquired below the line of text that is currently being read as readers apparently focus their attention only on the line being fixated.

Although the perceptual span extends about 14–15 letter spaces to the right of fixation, the area from which words can be identified on a given fixation (the word identification span) generally does not exceed 7–8 letter spaces to the right of fixation. However, neither the perceptual span nor the word identification span is fixed as both can be modulated by word length. For example, if three short words occur in succession, readers are able to identify all of them. If the upcoming word is highly constrained by the context, readers acquire more information from that word than from unpredictable words, and if the fixated word is difficult to process, readers obtain less information from the upcoming word.

Orthography also influences the size of the perceptual span. Hebrew readers have a smaller span than English readers, and Japanese and Chinese readers have even smaller spans. Hebrew is a more densely packed language than English, and Japanese and Chinese are both more densely packed than Hebrew; densely packed refers to the fact that it takes more letters per sentence in English than Hebrew, for example. Finally, reading skill influences the size of the perceptual span. Beginning readers (at the end of second grade) have a smaller span than skilled readers and adult dyslexic readers have smaller spans than skilled readers.

4. Integration Of Information Across Eye Fixations

It is clear from research using both the moving window and the boundary technique that there is a preview benefit from having the word to the right of fixation available on an eye fixation, and that information obtained about the parafoveal word on fixation n is combined with information on fixation n 1 to speed identification of the word when it is subsequently fixated. What kind of information is integrated across consecutive eye fixations in reading?

Experiments using the boundary paradigm described above have varied the orthographic, phonological, morphological, and semantic similarity between an initially displayed stimulus and a target word in attempts to determine the basis of the preview effect. A number of major findings have emerged from these studies. First, there is facilitation due to orthographic similarity as having a preview of chest facilitates the processing of chart. However, the facilitation is not strictly due to visual similarity because changing the case of letters from fixation to fixation (so that ChArT becomes cHaRt on the next) has little effect on reading behavior. Second, the facilitation is in part due to abstract letter codes associated with the first few letters of an unidentified parafoveal word; although there may be some facilitation from other parts of the word to the right of fixation, the bulk of the preview effect is due to the beginning letters. The effect is not simply due to spatial proximity because there is facilitation from the beginning letters of words when readers are asked to read sentences from right to left, but with letters within words printed from left to right. Third, there is facilitation due to phonological similarity, since beech facilitates beach and shoot facilitates chute, with less facilitation in the latter than in the former case. Fourth, although morphological factors can influence fixation time on a word, they are not the source of the preview benefit. Finally, there is no facilitation due to semantic similarity since song as the initial stimulus does not facilitate the processing of tune. It thus appears that orthographic (abstract letter) and phonological codes are the basis of the preview benefit and used in integrating information across fixations.

5. Eye Movement Control: Where To Fixate Next

There are two components to the issue of how eye movements are controlled during reading: (a) where to fixate next and (b) when to move the eyes. It appears that there are separate mechanisms involved in these decisions and they will accordingly be discussed separately.

Word length seems to be the primary determinant of where to fixate next in skilled reading. When length information about the upcoming word is not available, readers move their eyes a shorter distance than when such information is available. Also, the length of the word to the right of fixation strongly influences the size of the saccade. Furthermore, there is a landing position effect such that readers tend to fixate about halfway between the beginning and the middle of words. However, there is also a launch site effect such that where readers land in a word is strongly influenced by where the saccade came from. Thus, whereas the most frequent landing position is near the middle of the word, if the prior saccade was launched some distance (8–10 letters) from the target word then the landing position shifts to the left of center. Likewise, if the prior saccade was launched close (2–3 characters) to the beginning of the target word, the landing position shifts to the right of center. Given this, the optimal strategy would be to fixate near the middle of each successive word. However, since short words can often be identified when they are just to the right of fixation, they are often skipped. Factors such as these result in the landing position distribution being spread somewhat.

6. Eye Movement Control: When To Move The Eyes

What determines when we move our eyes? The amount of time a reader fixates on a word or segment of text reveals something about the cognitive processes associated with comprehending that word or segment. One reason for this is that information gets into the processing system very early in a fixation (thus leaving a lot of time for processes associated with word recognition and other necessary comprehension processes). Experiments using the foveal mask paradigm in which the onset of the mask is delayed following a saccade have demonstrated that reading proceeds quite normally if the reader has 50 ms to process the text prior to the onset of the mask. If the mask occurs earlier, reading is disrupted. Although readers typically acquire the visual information needed for reading during the first 50 ms of a fixation, they can extract information at other times during a fixation as needed.

There have been many demonstrations that various lexical, syntactic, and discourse factors influence fixation time on a word. In particular, all of the following variables influence fixation time on a word: (a) word frequency, (b) contextual constraint, (c) semantic relationships between words in a sentence, (d) anaphora and co-reference, (e) lexical ambiguity, and (f ) syntactic disambiguation. Thus, for example, readers look longer (the first fixation duration and gaze duration are both longer) at words that do not occur as frequently in the language than they look at more frequent words, and they look at words that are highly predictable for less time than they look at words that are not predictable from the preceding context. In essence, the amount of time that readers look at words is strongly influenced by the ease or difficulty associated with understanding that word. Other factors also influence how long readers look at words. When reading a sentence such as ‘While Mary was mending the sock fell off her lap’ (referred to as garden-path sentences), readers fixate for a long time on fell. This is because they typically parse the sock as a direct object and then don’t know what to do with fell when they get to it. Thus, they need to reanalyze the sentence. In this particular example, readers would not only fixate for a long time on the disambiguating word ( fell ), they would also make a regression back to the earlier part of the sentence where the misanalysis began.

In essence, the variability in fixation durations that was mentioned at the outset of this research paper is largely due to the ease or difficulty of processing the fixated words. Because the oculomotor system is apparently closely linked to the part of the brain responsible for processing language, eye movements are a very good way to study various questions related to language and comprehension. Some computational models of eye movements in reading (see Legge et al. 1997, Reichle et al. 1998) have been developed that nicely capture many of the properties of skilled reading.

Bibliography:

1. Crowder R G, Wagner R K 1992 The Psychology of Reading, 2nd edn. Oxford University Press, New York
2. Frazier L, Rayner K 1982 Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology 14: 178–210
3. Just M A, Carpenter P A 1980 A theory of reading: From eye fixations to comprehension. Psychological Review 87: 329–54
4. Just M A, Carpenter P A 1987 The Psychology of Reading and Language Comprehension. Allyn and Bacon, Boston
5. Legge G E, Klitz T S, Tjan B S 1997 Mr. Chips: An ideal observer model of reading. Psychological Review 104: 524–53
6. McConkie G W, Rayner K 1975 The span of the effective stimulus during a fixation in reading. Perception and Psychophysics 17: 578–86
7. Pollatsek A, Bolozky S, Well A D, Rayner K 1981 Asymmetries in the perceptual span for Israeli readers. Brain and Language 14: 174–80
8. Rayner K 1975 The perceptual span and peripheral cues in reading. Cognitive Psychology 7: 65–81
9. Rayner K 1998 Eye movements in reading and information processing: 20 years of research. Psychological Bulletin 124: 372–422
10. Rayner K, Bertera J H 1979 Reading without a fovea. Science 206: 468–9
11. Rayner K, Pollatsek A 1989 The Psychology of Reading. Prentice-Hall, Englewood Cliffs, NJ
12. Reichle E D, Pollatsek A, Fisher D L, Rayner K 1998 Toward a model of eye movement control in reading. Psychological Review 105: 125–57