orthography and the development of reading processes: an

Orthography and the Development of Reading Processes: An Eye-Movement

Study of Chinese and English

Gary FengDuke University

Kevin MillerUniversity of Michigan

Hua Shu and Houcan ZhangBeijing Normal University

As children become proficient readers, there are substantial changes in the eye movements that subservereading. Some of these changes reflect universal developmental factors while others may be specific to a par-ticular writing system. This study attempts to disentangle effects of universal and script-dependent factors bycomparing the development of eye movements of English and Chinese speakers. Third-grade (English: meanage = 9.1 years, n = 23; Chinese: mean age = 9.4 years, n = 25), fifth-grade (English: mean age = 11.2 years,n = 30; Chinese: mean age = 11.4, n = 25), and undergraduate students (English: n = 26; Chinese: n = 30) readstories in their native language while their eye movements were recorded. Results show a mixture of orthog-raphy-dependent factors with others that are remarkably parallel across these two very different writing sys-tems. Orthographic effects are also more pronounced for children than for skilled adult readers. Implicationsfor theories of reading eye movements and reading development are discussed.

As children become proficient readers, the mechan-ics of reading eye movements undergo substantialchanges. On average the duration of fixations (theperiod when the eyes stay relatively motionless)decreases, and the length of saccades (the jumpsbetween fixations) increases. Between them, thesechanges lead to a more than tripling of readingspeed between first and twelfth grade according toone estimate (Taylor, 1965). This is a remarkablechange, and it doubtless reflects increased readingexpertise and other developmental factors, as wellas features of the writing system being read. Thegoal of the present study was to delineate how thesefactors influence the moment-by-moment decisionof when and where to move one’s eyes in reading.

Two broad classes of factors might account forthese changes. The first involves universal improve-ment of efficiency in the course of reading acquisi-tion regardless of the language or writing system.Some of these changes may reflect maturationalprocesses of the brain, for example, the ability

to inhibit inappropriate saccadic eye movements(Fischer, Biscaldi, & Gezeck, 1997; Klein & Foerster,2001) or the remarkable increases in processingspeed that occur across many cognitive domains(Kail, 1986). In other cases, increasing reading facil-ity involves general factors that are inherent in theprocess of reading such as word recognition andsentence comprehension. A case in point is the per-ceptual span in reading, which is the region withinwhich useful visual information is perceived(Rayner, 1998). Rayner (1986) reported that theperceptual span expands with age and readingexpertise. These processes are experience relatedrather than maturational, but they reflect aspects ofreading that are so universal that experience inreading any orthography should lead to similarpatterns of increasing eye-movement efficiency.

The other class of factors that might account forincreasingly efficient eye-movement control withdeveloping reading skill involves efficient adaptationand exploitation of the features of a particularorthography. To the extent that writing systemsdiffer in their structure and fundamental units, onemight expect that the developmental course ofThis study was supported by National Institutes of Health

Grant 5R03HD051909-02 to the first author and by NationalScience Foundation Grant BCS 00-79973 to the second author.

Correspondence concerning this article should be addressed toGary Feng, Department of Psychology & Neuroscience, P.O. Box90085, Duke University, Durham, NC 27708. Electronic mail maybe sent to [email protected].

Child Development, May/June 2009, Volume 80, Number 3, Pages 720–735

� 2009, Copyright the Author(s)

Journal Compilation � 2009, Society for Research in Child Development, Inc.

All rights reserved. 0009-3920/2009/8003-0009

reading eye-movement control would differ amongchildren learning to read such different orthogra-phies. This would only be true to the extent that chil-dren’s reading processes are sensitive to the factorsrepresented by those orthographic differences, andso understanding the nature and timing of the effectsof orthography on reading will provide insights intothe developing flexibility of reading processes.

In an attempt to begin to distinguish contributionsof universal and orthography-dependent factors toreading development, the present study compareseye movements of children and adults who arespeakers of two very different languages Englishand Chinese. Because the orthographies used towrite these two languages differ so greatly, compari-sons of the developmental course of reading inChinese and English provide a good basis for begin-ning to disentangle script-dependent aspects ofreading development from those that rely on moreuniversal aspects of reading and its acquisition.

Before introducing the design of the study, wewill first review existing data on developmentalchanges in reading eye movements of readers ofEnglish, then discuss features of the English andChinese orthographies relevant to reading, andfinally discuss existing eye movement data on read-ing in these two orthographies.

Eye Movement Control in Reading

During reading, the eye alternates between jerkymovements called saccades and brief stops calledfixations. The primary function of eye movements isto bring new words to the retina because what wecan perceive during a fixation is limited by the per-ceptual span, which is typically less than 20 lettersfor English reading adults (see Rayner, 1998). Thisis a challenging task. The decision of where to sendthe eye has to be made very rapidly (typicallyunder 200 ms) with limited visual information.And when the move turns out to be nonoptimal,the reader is stuck with the result until the next sac-cade. In some sense, eye movements are a bottle-neck of information processing in reading, that is,they control the speed and order of informationintake. Thus, efficient eye movement planning isnot only a reflection of reading expertise, but also acritical part of the reading apparatus.

There are two important decisions in reading eyemovement planning: (a) where to send the eye and(b) when to terminate the present fixation and movethe eye. There is mounting evidence that they aremade by two separable but overlapping neuralpathways, particularly in the superior colliculus

and the brain stem (Findlay & Walker, 1999). Feng(2006a) also showed that the two pathways can besimulated as separate Markov processes in reading.

This raises the possibility that various aspects ofreading eye movements may be differentially sensi-tive to maturational and experiential changes.There is some suggestive evidence from nonreadingoculomotor studies. Fischer et al. (1997) showedthat the oculomotor processes for voluntary controlof saccades (initiating saccades and inhibiting in-appropriate responses) correlate with age, but thosefor maintaining fixations do not. The former hasbeen linked with prefrontal maturation (Klein &Foerster, 2001), whereas the latter is mostly a func-tion of the superior colliculus and other vision-related centers. Voluntary control is involved inchoosing the next word target, although it may alsoplay a role in determining the duration of fixations.Thus these results suggest strong developmentalchanges in the where pathway in reading, althoughthe specific orthography may or may not have animpact on the where pathway. In addition, someaspects of the when decision may be controlled byrobust neurological processes that are not influ-enced by either reading experience or age.

Developmental Changes in Reading Eye Movements

Much of what we know about reading eye move-ments comes from studies of English-speakingadults (see Rayner, 1998); studies on the normativedevelopment of reading eye movements remainfew and far between (see Buswell, 1922; Judd et al.,1918; McConkie et al., 1991; Rayner, 1986; Taylor,1965). In general, results show that the efficiency ofeye movements increases as a function of age andreading proficiency. For example, McConkie et al.(1991) found that the average fixation durationdecreased from 304 ms in first grade to 262 ms inthird grade and then to 243 ms in fifth grade, andthe mean saccade length increased from 3.6 lettersto 5.7 and to 6.3 letters, respectively. In addition,children change their scanning patterns. The num-ber of fixations per word drops continuously fromfirst through sixth grade (Buswell, 1922; Rayner,1986, 1998; Taylor, 1965). This is largely driven by areduction of within-word refixations and anincrease of word skipping (McConkie et al., 1991),both of which are signs of proficient reading.

Developmental studies also hint that someaspects of eye movements may show little changewith age. Three examples will be briefly discussedhere. The first is the landing position distribution,or where the eye looks within a word (McConkie,

Reading Development 721

Kerr, Reddix, & Zola, 1988). Adult readers gener-ally target the center of the word because word rec-ognition is most efficient there. First-grade studentsshow almost identical landing position distribu-tions to those of skilled readers (McConkie et al.,1991), suggesting that this low-level oculomotoroptimization is already in place when childrenbegin to read. Second, the mechanism that controlsfixation duration is also less susceptible to develop-mental forces. Although mean fixation durationdeclines with age, the mode (peak) of the fixationduration distribution remains nearly constantacross age groups, at around 180 ms (McConkieet al., 1991). Young readers have longer mean fixa-tion durations primarily because of higher propor-tions of long fixations, which are often associatedwith cognitive control and inhibitory processes(Yang & McConkie, 2001). Finally, across a numberof studies, the proportion of regressive eye move-ments has been found to remain relatively flatthroughout elementary years (see McConkie et al.,1991); the regression rate varies between 20% and33%, depending on the studies. There is currentlyno convincing explanation for this phenomenon.

In summary, there is converging evidence thataspects of the oculomotor system may show a dif-ferent developmental course. One question impossi-ble to answer from studying readers of alphabeticscripts is the extent to which the development ofeye-movement control processes are affected byfeatures of the writing system children are learningto read. By comparing the development of childrenwho are learning to read very different orthogra-phies, we hope to begin the process of understand-ing the nature, timing, and limits of influences oforthography on the one hand, and maturation andgeneral reading experience on the other, on thedevelopment of eye-movement control processesthat subserve skilled reading.

Chinese and English Orthographies: Features Relevantto Reading

Chinese characters and the alphabetic orthogra-phy used to write English differ in a number offeatures with the potential to affect reading acquisi-tion. The main features relevant here are: (a) thefundamental linguistic unit represented by theorthography, (b) the relative transparency of theorthography, and (c) the orthographic marking ofboundaries between units. These features interactto produce important differences in the perceptibleinformation available to readers of these twoorthographies.

Fundamental orthographic units. English wordsare composed of a series of letters, which veryroughly correspond to phonemes. This is not to saythat English letters correspond to phonemes in anystraightforward way (Treiman, Weatherston, &Berch, 1994). With very rare exceptions, Chinesecharacters represent morphemes, and correspondto syllables in the spoken language. Characters can-not be further decomposed into smaller phonologi-cal units. Thus, the two scripts differ in thephonological transparency, i.e., the likelihood that areader can pronounce an unfamiliar word.

Transparency of writing systems. Both writingsystems encode phonological and morphologicalinformation, but with distinct approaches. Thetype and consistency of information afforded bythe orthography determine the efficacy of begin-ning readers’ decoding strategy (Seymour, Aro, &Erskine, 2003).

The Chinese writing system provides hints toboth the sound and meaning of a character, thoughtheir usefulness depends on the character and thereader. About 90% of Chinese characters aresemantic–phonetic compounds (Hoosain, 1991),consisting of one part that has some relation to thepronunciation of the character and one part thathas some relation to its meaning. One would havea better chance guessing an unfamiliar semantic-phonetic compound character than other types ofcharacters. Nevertheless, the proportion of thesecharacters is smaller in early grades; that is, itaccounts for 45% of characters introduced in thefirst grade compared with 86% in the fifth grade(Shu, Chen, Anderson, Wu, & Xuan, 2003). Phono-logically consistent characters, characters that hadthe same pronunciation as their phonetic compo-nents or differed only by tone, rose from 32% in thefirst grade to 43% in the sixth grade. Thus, the mor-phology of Chinese characters provides useful cluesto their pronunciation, but not in a way that chil-dren can expect to be able to ‘‘sound out’’ a newcharacter. This is particularly the case for youngchildren, who are more likely to be learning charac-ters whose pronunciation cannot be derived from aphonetic component; even if it does, young readersmay not know the character that serves as a pho-netic component in a new character. There are otherways that characters are formed. For example, asmall percentage of characters (< 1% according toDeFrancis, 1989) have their origin in pictographicrepresentations of a concept. Because these charac-ters have become conventionalized over time, themeaning of such characters is rarely evident toreaders.

722 Feng, Miller, Shu, and Zhang

As is the case with Chinese, English containsmany exception words whose pronunciation cannotbe inferred from their written representations(Nagy, Anderson, & Herman, 1987). But in general,a skilled reader has a good chance of guessingmost, if not all, sounds of an unfamiliar Englishword (Venezky, 1999). The greater likelihood ofbeing able to figure out new words for English mayhave consequences for eye movement control. Forexample, readers of English may be more likelyto gain new information by repeatedly inspecting(refixating) a word, whereas the same strategy ismuch less likely to pay off in reading Chinese,particularly for young readers.

Orthographic marking of word boundaries. A majordifference between the two orthographic systems isthat English texts consist of a series of wordsseparated by spaces, whereas Chinese texts consistof a series of evenly spaced characters that largelycorrespond to morphemes. Words in modern Chi-nese consist of one or more characters (with anaverage of 1.5 characters; Feng, 2006b; Sun, Morita,& Stark, 1985) but they are not marked in any formin writing. Consequently, there is little consensusamong Chinese speakers about what words are(Miller, Chen, & Zhang, 2007; Wang, 2003).

The difference in word vs. morpheme spacingmay have both conceptual and perceptual conse-quences (Feng, 2008). If words are essential in thecomprehension process, one would predict someprocessing costs associated with assembling charac-ters into words in reading Chinese, and these costsshould be larger for beginning readers. The lack ofword spacing may also lead to costs in visualperception and oculomotor planning. Skilled read-ers of English acquire word-length informationparafoveally and use it in deciding where to looknext (Morris, Rayner, & Pollatsek, 1990; Rayner,Sereno, & Raney, 1996). The lack of word boundarydemarcations in Chinese would preclude theseuseful oculomotor strategies and could potentiallyslow down Chinese reading.

Interestingly, the opposite prediction may alsobe made. Past research on Chinese reading hasshown that adding spaces to separate wordsresulted in either no effect (Everson, 1986) or nega-tive effects on comprehension (Liu, Yeh, Wang, &Chang, 1974; see Feng, 2006b, for a review). Thus itmay be that skilled readers of Chinese can organizecharacters (morphemes) into words without takingresources away from reading comprehension.Furthermore, there may be a trade-off betweenmarking word boundaries (as in English) versusmorpheme boundaries (as in Chinese). Although

word spacing may make saccade programming arelatively straightforward task in English reading,it may impose costs as well as benefits becausemorphological boundaries are not marked (Reichle& Perfetti, 2003).

There is some evidence supporting this possibil-ity. Long words are more likely to be refixated, andthis may reflect some kind of morphological pro-cessing in which words are broken into smallerunits. German is a language notorious for its longwords (Twain, 1880 ⁄ 1997). Inhoff, Radach, andHeller (2000) showed when spaces are inserted tobreak long words into smaller, meaningful units,native German readers performed better on namingand reading tasks. Evidence of morphological pro-cessing costs has also been reported in Finnish andEnglish compound word reading (Andrews, Miller,& Rayner, 2004; Bertram & Hyona, 2003; Hyona,Bertram, & Pollatsek, 2004). These studies involveskilled readers only; whether and when sensitivityto the unit marked by an orthographic systemdevelops will be explored in this study.

Reading in Chinese and English: Eye MovementPatterns

Given the large differences between the writingsystems, one might predict distinct patterns of eyemovements in reading Chinese and English.Research in the last 80 years, however, suggeststhat there are more similarities than differencesamong skilled readers of these two orthographies(Feng, 2006b; Feng, Miller, Shu, & Zhang, 2001;Gray, 1956; Hsien-Ming Yang & McConkie, 1999;Miles & Shen, 1925; Peng, Orchard, & Stern, 1983;Sun et al., 1985; Tsai & McConkie, 1995, 2003;Yang, 1994; Yang & McConkie, 1994). For example,when native Chinese or English speakers read com-parable scientific articles (Sun & Feng, 1999), criticaleye movement variables, such as mean fixationdurations (257 ms vs. 265 m), average saccadelength (1.71 vs. 1.75 words per fixation), and read-ing rate (386 vs. 382 words per minute) wereremarkably similar. The perceptual span, the size ofeffective vision during a fixation (McConkie &Rayner, 1975) is also comparable among adult read-ers of Chinese and English when measured inwords, not letter spaces (Inhoff & Liu, 1998; Tsai &McConkie, 1995; Tsai, Tzeng, Hung, & Yen, 2000).

Eye-movement studies of Chinese reading arestill quite rare. Chen, Song, Lau, Wong, and Tang(2003) reported data involving second-, fourth-, andsixth-grade students from Hong Kong readingsixth-grade-level prose passages. Mean first fixation


duration was 280, 253, and 232 ms, respectively, forthe three grades. Average forward saccade lengthincreased from 2.3 character spaces for the secondgraders to 2.6 for the older grades. The probabilityof regression remained virtually flat at around15%–17%. A decrease in the rate of within-wordrefixation was hinted by the large decline in gazeduration (sum of duration of all fixations on a wordbefore the eye leaves the word). Overall, the devel-opmental pattern reported in Chen et al. is broadlyconsistent with the literature on English readingdevelopment summarized before.

Although research looking at college students’reading of Chinese and English has been notablefor the cross-language consistency found in themechanics of reading, this need not imply that thesame patterns will be found for children. With lessreading experience and more fragile skills, youngchildren may be more affected by the characteristicsof orthography than are adults. To our knowledge,the study reported here is the first cross-languagedevelopmental study that compares eye move-ments of Chinese and English speakers.

Predictions

Studies of the development of reading pro-cesses in children learning to read Chinese andEnglish provide a powerful basis for disentan-gling the role of universal and script-dependentprocesses in the development of reading skill.Logically, there are three possible patterns ofdevelopment among the different components ofreading eye-movements:

Parallel development across writing systems. To theextent that some aspects of eye-movement controlshow parallel development across two very differ-ent writing systems, developmental explanationsthat refer to global cognitive changes or universalaspects of the reading process are implicated. Thisis a plausible hypothesis for at least two reasons.First, maturational processes have been shown toplay important roles in oculomotor development(Fischer et al., 1997). Second, existing studiessuggest that reading Chinese is remarkably similarto reading English among skilled readers of thetwo orthographies (e.g., Gray, 1956; Sun & Feng,1999). Whether the same pattern holds acrossdevelopment is a question to be addressed in thisresearch.

Script-dependent developmental patterns. As des-cribed above, Chinese characters and Englishalphabetic writing differ profoundly in theirvisual organization and in the way in which they

represent their respective spoken languages. Thesedifferences may impose different demands on theoculomotor system and result in fundamentallydifferent reading processes. Specifically, because(a) information is more densely packed inChinese characters compared with English words(Hoosain, 1991), (b) characters provide lesssystematic phonological information, and (c) Chi-nese words are not separated by spaces, onemight predict that Chinese children should tendto make shorter saccades and longer fixations onaverage, compared with their American peers.Together, these should lead to slower readingspeeds among Chinese children.

But one might also make the opposite prediction.Chinese-speaking children may have developedscript-specific oculomotor strategies early on, andthus not be negatively affected by the lack of wordspacing. In addition, they may benefit from not hav-ing to break long words into smaller meaning units.If breaking up is indeed hard to do, then it shouldbe readers of English who show disruption becauseof the need to parse long perceptual units in thecourse of reading. This disruption should show upin relatively more refixations of words among read-ers of English compared with their Chinese peers.

Different pathways for different aspects of eye-move-ment control. The development of reading eyemovements may be driven by a combination ofscript-dependent features and others that reflectmaturation or aspects of reading that do not varyacross writing systems. Fischer et al. (1997) distin-guished between neural systems that control whereone looks and when one changes looking position.Based on prior research with children learning toread English, it is possible that these systems mightshow different patterns of variation across orthog-raphies, with orthography affecting where childrenlook more than how long children look during fixa-tions.

Differential cross-orthography developmentalpatterns would provide strong support for the dif-ferentiation between these two eye-movement con-trol systems. This model leads to the followingpredictions: Fixation duration should show similardevelopmental trajectories across age (McConkieet al., 1991) and languages (Sun & Feng, 1999). Sac-cade planning, on the other hand, may be moresusceptible to cognitive control (such as differentreading strategies) as well as to influences of idio-syncratic features of writing systems. We predictthat saccade-related measures will demonstrate (a)substantial developmental changes and (b) clearcross-language differences.


Method

Participants

The American participants were 23 third-gradestudents (mean age = 9.1 years, range = 8.6–10.2; 12males and 11 females), 30 fifth-grade students(mean age = 11.2 years, range = 10.6–12.1; 15 malesand 15 females), and 26 undergraduate students (14males and 12 females) from two small towns in eastcentral Illinois. They were primarily Caucasians. Asa group, children from the two Illinois schoolstested above the 80th percentile in the Illinois Statereading assessments. The Chinese participants were25 third-grade students (mean age = 9.4 years,range = 9.0–10.6; 12 males, 13 females), 25 fifth-grade students (mean age = 11.4, range = 10.7–11.9;12 males and 13 females), and 30 undergraduatestudents (14 males and 16 females) from Beijing,China. The Chinese school is located near down-town Beijing. At the time of the study, it was wellequipped but was rated as average in academicperformance in its district. All participants in bothcountries had either normal vision or corrected tonormal vision.

Materials

The selection of reading materials was subject toconflicting constraints. On the one hand, theyshould be age and culture-specific and reflect mate-rials children might encounter in their daily lives; aone-size-fit-all approach could introduce unwantedage and cultural biases. On the other hand, a com-mon metric is required to compare reading eyemovement patterns across languages and age. Thecompromise was a mixed design (see Appendix Afor more details).

To compare developmental changes in readingeye movements on a common metric, a core subsetof materials was read by all readers in their ownlanguages. This included two passages ‘‘ThePower of Boats’’ (BOAT) and ‘‘The Foolish Mule’’(MULE) that had parallel English and Chineseversions and were within the reading abilities ofboth third- and fifth-grade participants. The twostories and the comprehension questions were orig-inally used in a cross-cultural study of American,Japanese, and Chinese students’ reading abilities(Stevenson et al., 1990). Cultural biases and diffi-culty levels of these stories were controlled in thedevelopment of these materials. They were rated ashaving fourth-grade and fifth-grade difficultylevels, respectively, in Stevenson et al.’s (1990)study, but our pretests and teacher interviews

showed that most third graders had no difficultywith them. These two stories served as anchors forcross-cultural and cross-age comparisons.

An equally important concern was that readingmaterials should reflect the cultural value and dailyexperience of readers, particularly young readers,of the two countries. To this end, different materialswere selected for each grade and language. We didthis by first obtaining a selection of stories frompopular third- and fifth-grade extracurricular read-ing series in each country. We selected two storiesfrom the third-grade series and two from the fifth-grade series for each country that were roughlymatched in terms of their contents (e.g., biography,science, etc.) and approximate length. Teachers ofthe participating classes read the passages andjudged them to be appropriate for their students.As a check on the difficulty of the stories, we askedteachers in both countries to circle words (English)or characters (Chinese), that they thought would bedifficult for an ‘‘average’’ student in their classes.On average, the Chinese materials containedslightly more unfamiliar characters (25 for thirdgrade and 6 for fifth grade) compared with unfa-miliar words in the English materials (17 for thirdgrade and 7 for fifth grade). These passages providea sampling of the kinds of reading that children ineach country might be naturally exposed to, butnone of our participants had read the stories priorto the experiment. We generated two comprehen-sion questions for each story.

Finally, we asked adult readers to read the samestories as young readers did. The adult experimentswere not designed to measure how skilled readersread reading-level appropriate materials. The focusof the present study is the nature of children’s read-ing and its development. We included adults ineach country primarily to get a baseline measure ofhow skilled readers would read the kinds of mate-rials we were showing to children. As a result,adult readers’ reading and eye movement behav-iors may differ from when they read more sophisti-cated materials.

Reading materials were presented on a computerscreen 60–70 cm away from the reader. Childrenread the stories selected for their age group; adultsread all of the stories. The Chinese materials weredisplayed in 24 · 24 pixel Song font, correspondingto about 1.3 visual degrees per character with 4pixels space between characters. An English letterwas on average 7.3 pixels wide or 0.35 visualdegrees. Each screen held a maximum of sevenlines of English text (45 pixels between lines) or sixlines of Chinese text (48 pixels between lines). The


line spaces were adequate to allow for typical verti-cal drifts of the Eyelink eye tracker.

Font sizes and line spaces were chosen to facili-tate comparison of eye movement parametersbetween the two languages. Cross-linguistic com-parisons of saccade length are complicated by thefact that there is not a common metric for Englishand Chinese texts. Because reading saccades areprimarily affected by linguistic units (number ofletters or words) rather than by font size and view-ing distance (Rayner, 1998), saccade length is typi-cally reported in letter spaces for English readingbut in character spaces for Chinese reading. Arbi-trary scaling factors have been proposed, such astwo English letters for a Chinese character (Yang &McConkie, 1994), but their appropriateness is ques-tionable (e.g., Feng, 2006b; Hoosain, 1991). In thepresent study, we adjusted the font size in bothlanguages so that a parallel story in English andChinese (i.e., the MULE story) would have approxi-mately the same number of lines in both languages.As a result, a Chinese character equals 3.9 Englishletter spaces. A typical six-letter English wordtherefore took approximately the same space as theaverage length of a Chinese word, which is 1.5characters (Hoosain, 1991; Tsai, Lee, Hung, &Tzeng, 2001). This was obviously a crude attemptto equate the materials and was never intended tobe a precise matching. Nevertheless, it allowed acomparison of saccade length in term of the screenpixels. Although any cross-language comparison ofsaccade-length will be problematic, developmentalchanges within country do not have these concerns.

As noted previously, the Chinese orthographydoes not mark word boundaries, and educated Chi-nese-speaking adults do not necessarily agree onhow to divide a passage into words (Hoosain, 1991;Miller et al., 2007; Wang, 2003). We identifiedwords based on a word-segmentation protocolknown as the Chinese National Standard GB13715(Liu, Tan, & Shen, 1993). Statistics and analyzesinvolving word units in Chinese are all based onthis standard.

Apparatus and Procedure

Participants were asked to read the storiessilently as they ordinarily would and were informedthat they would be asked some questions abouteach story after they completed reading it. Readers’eye movements were recorded with an EyeLink Ieye-tracking system, which is a head-mountedinfrared system with 250-Hz sampling rate and amaximum spatial resolution of 0.005�, although the

typical accuracy level is approximately 0.5 visualdegree, measured by repeated calibrations. Saccadedetection was based on a built-in acceleration-basedalgorithm with default parameters for reading andcognitive studies. No head support or chin rest wasused. Head movements were corrected by the built-in algorithm in the eye tracker.

The English and Chinese studies were both runby the first author and the procedure was identical.A 9-point calibration was performed before eachexperiment and was repeated as necessary duringthe experiment. A calibration was accepted whenaverage error over the 9 points was less than 0.8degrees. A 1-point drift correction was performedbetween each page. Comprehension questions werepresented after each story and were answeredorally. The stories were presented in a fixed orderin each language. Reading typically took approxi-mately 10–20 min and participants could ask for abreak at any time, which several children did butnot the adults.

Results

Reading Comprehension

The rates of correct answers to the comprehen-sion questions were on average 67.8%, 89.2%, and97.2% for American third graders, fifth graders,and undergraduate students, respectively, andwere 78.0%, 81.8%, and 94.4% for Chinese thirdgraders, fifth graders, and undergraduate students,respectively. Because the reading passages andcomprehension questions were different in the twolanguages, these numbers cannot be compareddirectly. Comparing comprehension levels for theBOAT and the MULE stories is more straightfor-ward, because the stories and questions are compa-rable across languages. The mean score for the fourquestions on these two stories showed no signifi-cant country difference between Chinese andAmerican readers, F(1, 163) = 3.304, p = .071, but asignificant developmental trend, F(2, 163) = 11.619,p < .001, (69.3%, 83.1%, and 90.4%, for Americanthird graders, fifth graders, and adults; 78.0%,88.0%, and 97.7% for Chinese third graders, fifthgraders, and adults, respectively).

Reading Speed

One prediction based on differences betweenthe English and Chinese writing systems is thatthere may be more processing costs associatedwith reading Chinese, particularly for Chinese


children. If this is true, one would expect slowerreading and longer processing time per word. Fig-ure 1 shows the average reading time per word,averaged across all stories; a separate analysis withonly the BOAT and the MULE stories showed thesame pattern and therefore is not reported sepa-rately. Reading time per word decreased with age,F(2, 148) = 83.46.14, p < .001, but there was a sig-nificant interaction between age and language, F(2,148) = 3.36, p = .037. Paired comparisons withineach age group showed that Chinese children readsignificantly or marginally significantly faster thantheir American counterparts, F(1, 148) = 4.95,p = .028 in third grade, F(1, 148) = 3.85, p = .051 infifth grade, but the reading time for American andChinese undergraduate students did not differ sig-nificantly, F(1, 148) < 1. American readers’ readingspeeds, when converted into words per minute(wpm), were approximately 110, 140, and 340wpm for third grade, fifth grade, and undergradu-ate students, respectively. The figures for youngreaders are consistent with figures reported inprior literature. At 340 wpm, the reading speed ofadult English speakers was faster than averagesreported in the literature, approximately 300 wpm(Carver, 1990; Rayner, 1998). However, as Carver(1990) pointed out, a direct comparison of thereading rate may be misleading when materialsdiffer in word length. The average word length inthis study (4.03 letters per word) is much shorterthan that in adult materials (typically six lettersper word; Carver, 1990), which is likely to haveinflated the reading rate. Other indicators such asthe level of comprehension suggest that adultreaders were engaged in meaningful readingbehaviors. We were unable to find reliable norms

of reading rate in Chinese for either children orskilled readers.

As we expected, although English and Chinesespeaking adults read at a similar speed, largelanguage differences were found among beginningreaders. The direction of the differences, however,is opposite to the prediction laid out earlier, withan advantage for young Chinese readers despitethe lack of explicitly marked word boundaries intheir orthography. In the following analyses,we decompose story reading time into twofactors—number of fixations per word and meanfixation duration—and explore their relation toreading speed.

Number of Fixations

The average number of fixations per word was2.00, 1.56, and 0.79 for American third graders,fifth graders, and adults, and 1.71, 1.31, and 0.86for their Chinese counterparts, respectively. Asshown in Figure 2, the number of fixations perword decreased with age, F(2, 148) = 94.30,p < .001, and was lower for Chinese than for Eng-lish readers, F(1, 148) = 6.61, p = .011. There wasagain a significant interaction between age andlanguage, F(2, 148) = 3.68, p = .028. The differenceswere significant at third grade, F(1, 148) = 6.52,p = .012, and at fifth grade, F(1, 148) = 6.02,p = .015, but nonsignificant for adult English andChinese readers, F(1, 148) < 1. The same patternwas found when the BOAT and MULE storieswere analyzed separately.

Figure 1. Mean reading time per word.

Figure 2. Number of fixations per word.


Fixation Duration

Mean fixation duration was approximately 263,244, and 191 ms for American third-grade, fifth-grade, and adult readers, and 265, 238, and212 ms for Chinese third-grade, fifth-grade, andadult readers, respectively. Figure 3 shows theaverage duration of fixations of each group ofparticipants. Mean fixation duration decreasedsignificantly with age, F(2, 148) = 41.27, p < .001.There was no significant language main effect,F(1, 148) < 1, nor any interaction between ageand language, F(2, 148) = 2.06, p = .131. Averagefixation duration did not differ significantlybetween American and Chinese children, Fs(1,148) < 1, for both third and fifth grade, whereasAmerican college students’ fixation duration wasabout 21 ms faster than that of Chinese collegestudents, F(1, 148) = 4.89, p = .029.

Figure 4 compares the frequency distributionsof fixation duration for different reader groups.Only forward fixations were plotted, as regressivefixations are thought to have different functionsthan acquiring new information (Rayner, 1998).Three features are noteworthy in Figure 4. First,there is minimal developmental or cross-culturaldifference in the peaks of the distributions. All fix-ation duration distributions peak around 180 ms,despite differences in the means. Second, childrenin both languages differ from adult readers pri-marily in the right tail of the distribution, that is,children produce higher proportions of long fixa-tions and correspondingly lower percentages ofshort fixations than adults. Third, this phenome-non (more long fixations among children thanadults) is noticeably more pronounced among U.S.than Chinese children. Together, these findings

reinforce the conclusion of the relative lack of cul-tural differences in fixation duration reported ear-lier. They also add important qualifications to thepattern of mean fixation duration. As childrenbecome more proficient in reading, there is areduction in very long fixations, a phenomenonmore pronounced among those learning to readEnglish.

Distributions of Saccade Length

As mentioned previously, comparing eyemovement parameters is complicated by the dif-ferent unit of writing in English and Chinese.Our approach, which was far from the only onepossible, was to try to ensure texts of the samestory occupy approximately the same linearspace, in terms of pixels. Thus, the stories takeup the same horizontal space in total, and ana-lyzes of saccades can be based on pixels. We willalso report a word-based analysis of saccades.Because the texts are of the same size, effects ofthe different orthographies would show up in dif-ferent patterns of distributions of saccades amongforward fixations, regressions, and within-wordrefixations.

On average, American readers’ forward saccadeswere significantly shorter than those of Chinesereaders. The length of forward saccades increasedwith age, F(2, 148) = 81.82, p < .001, and was longerfor Chinese than for English readers, F(1,148) = 53.24, p < .001, and there was no interactionbetween age and language, F(2, 148) < 1. Therewere significant country differences for each agegroup: F(1, 148) = 13.41, p < .001 at third grade;F(1, 148) = 24.53, p < .001 at fifth grade; and

Figure 3. Mean fixation duration for American and Chinesereaders. Figure 4. Frequency distributions of fixation durations.


F(1, 148) = 16.64, p < .001 for adults. Mean saccadelengths for American readers were approximately5.5, 6.7, and 10.3 letter spaces for third-grade, fifth-grade, and adult readers, respectively. Saccadelength averaged 2, 2.4, and 3.1 characters for Chi-nese third-grade, fifth-grade, and undergraduatestudents, respectively.

Frequency distributions of forward saccadelength (in pixels) for American and Chinese chil-dren and adults are shown in Figure 5. The peaksof English and Chinese adult readers coincide, pos-sibly because of our effort to equate the physicallength. But the distributions differ otherwise, withAmerican children making more short saccadesand fewer long saccades than their Chinese peers.Because a typical five-letter English word isapproximately 37 pixels long, many of the shortsaccades made by American children would landon the same words. American undergraduate stu-dents, while still making fewer long saccades, wereno more likely to make these short saccades thanthe Chinese counterparts.

Kinds of Saccades: Forward, Refixations, andRegressions

The large number of short saccades in Americanchildren’s saccade distributions suggests that theymade more within-word refixations. To test thishypothesis, we divided fixations into three catego-ries—progressive fixations, interword regressions,and refixations—based on whether they landed on anew word, a word that had been read, or the currentword, respectively. The three categories are definedbased on words, for example, a leftward saccadethat landed on the same word—not infrequent inreading—counts as a refixation. Note that Chinesedata were also analyzed by words, as defined in theNational Standard GB13715, rather than charactersin this analysis. The percentages of each categorywithin each group are shown in Figures 6A–C. As isapparent from the figures, different patterns ofcountry and developmental effects were found forthe different kinds of saccades.

Progressive fixations. Progressive fixations incre-ased significantly with age, F(2, 148) = 166.79,p < .001, and were higher for Chinese than for Eng-lish readers, F(1, 148) = 10.70, p = .001, and therewas a significant interaction between age and lan-guage, F(2, 148) = 6.49, p = .002. The proportions offorward saccades did not differ in adult readers,F(1, 148) = 1.06, p = .305, but differed significantlyacross languages at third grade, F(1, 148) = 11.16,p = .001, and fifth grade, F(1, 148) = 9.88, p = .002.Thus the difference in reading speed reflects thegreater proportion of saccades among young Chi-nese readers that moved forward in the text.

Regressive fixations. A different pattern wasfound for regressions. There was a significant coun-try effect F(1.148) = 17.67, p < .001, but not signifi-cant age difference nor a significant interaction. At

Figure 5. Frequency distributions of saccade length.

Figure 6. Proportions of progressive fixations, regressive fixations, and refixations.


all ages, Chinese readers showed more regressionsthan did American participants, F(1, 148) = 6.126,p = .014 for third grade; F(1, 148) = 4.12, p = .044for fifth grade; and F(1, 148) = 7.79, p = .006 foradults. However, the magnitude of the effect isrelatively small compared with other differencesin Figure 6. Given that the physical lengths ofregressive saccades are similar between the twolanguages (see Figure 5), we suspect that the cross-linguistic difference may be related to the NationalStandard GB13715, which tends to aggressivelybreak words into smaller units. For the samesaccade length, smaller words will entail higherproportion of between-word regressions. Until abetter, psychologically more intuitive word parseris available for Chinese, the present result oflanguage differences in regression rate should betaken with a grain of salt.

We do not have a ready explanation for thelack of developmental changes. However, we arenot alone in finding this pattern. Data summa-rized in McConkie et al. (1991) showed thatregression rate remains flat throughout elementarygrades; Chen et al. (2003) also found that theregression rate remained at about 15% among Chi-nese readers ranging from second grade to collegestudents. The causes of regression in ordinaryreading and its relation to reading skill remainunclear, both from this study and from priorresearch.

Within-word refixations. Refixations decreasedwith age, F(2, 148) = 161.81, p < .001. There wasalso a significant language effect, F(1,148) = 41.97,p < .001, and a significant Age · Language interac-tion, F(2,148) = 5.67, p = .004. American children,but not American undergraduate students, madesignificantly more refixations than their Chinesepeers, F(1, 148) = 27.14, p < .001 for third grade;F(1, 148) = 21.82, p < .001 for fifth grade; and F(1,148) = 1.31, p = .253 for adults.

There are at least two plausible explanationsfor the greater incidence of refixations amongyoung readers of English compared with childrenlearning to read Chinese. The first involves thepossibility of sounding-out words in Englishversus Chinese, and the second involves theproblem of parsing English words into smallermorphologically meaningful units. A strategy ofrefixating unfamiliar words as part of an explicit‘‘sounding out’’ strategy is certain to be morefruitful in English than in Chinese. This explana-tion is supported by that fact that we did notfind significant differences in refixations betweenour samples of college students, who should not

need to sound out words in the children’s textsused in this study.

Because Chinese and English differ in the sizeof the fundamental unit of writing, the differencesin within-word refixations may also reflect prob-lems in parsing words into morphologically mean-ingful units. Supporting this explanation are theresults of Inhoff et al. (2000) who found that seg-mentation of long German words into smallerunits facilitated some (but not all) aspects of read-ing, even though these smaller words were unfa-miliar to their German college student sample.Because we chose to focus on normal text, whereword length, word frequency, and regularity arehighly confounded (Kliegl, Olson, & Davidson,1982), the present study cannot distinguishbetween these two explanations. Testing thesehypotheses will require specially designed materi-als in carefully controlled contexts. It is, indeed,possible that both factors play a role.

The different patterns of developmental trajec-tories and cross-language differences for thesedifferent components of eye movement controldo provide good evidence of the real but limitedways in which the eye movement systems tunethemselves to the features of particular orthogra-phies.

Discussion

To our knowledge, this is the first developmentalstudy that directly compares reading eye move-ments of English- and Chinese-speaking chil-dren. We went to great lengths to ensure theinternal and external validity of this cross-linguis-tic, cross-cultural development study. This isreflected in our sampling decisions, material con-struction, and measures taken to ensure consis-tency and comparability of the data collectedfrom the two very different cultures. The presentstudy focuses on normative development in read-ing naturalistic texts; this design does not affordtesting certain interesting hypotheses on linguisticprocessing or exploring individual differences ingreat depth. Adult eye movement parametersalso reflect the easy reading materials. Limita-tions notwithstanding, we believe that this dataset allows us to ask a number of important ques-tions, and in some cases, answer them withconfidence.

In what follows, we will first compare our resultswith published data, and summarize the develop-ment of reading eye movements in the context oftwo very different writing systems. We end with anoutline of a model to account for the diverse


developmental and cross-linguistic findingsreported here.

Cross-Linguistic Validity

One of the major challenges for any cross-cul-tural studies is to ensure the representativeness ofthe sample of materials and participants. We haveseveral reasons to believe that we were reasonablyclose to this goal.

Our English language data are in line with pre-vious eye movement studies. For example, themean fixation duration for American third- andfifth-grade students were 263 and 244 ms, respec-tively. These figures were almost identical to thosereported in McConkie et al. (1991), which were 262and 243 ms, respectively. Similarly, the averagesaccade length for third- and fifth-grade studentswas 5.5, and 6.7 letters in the present study versus5.7 and 6.3 letters in McConkie et al. (1991), respec-tively. Basic parameters for English-speakingadult eye movements fall in the typical range,although the slightly shorter mean fixation dura-tion (191 ms) and longer saccade length (10.3letters) reflected the fact that these stories, whichwere intended for third- and fifth-grade students,were very easy for adult readers. Chinese adults’eye movement parameters also align well withexisting data (see Feng, 2006b). Although our sam-ples of Chinese-speaking children differ in gradesfrom those in Chen et al. (2003) study (Grades 3and 5 in our study vs. Grades 2, 4, and 6 in Chenet al., 2003), mean fixation durations from the twostudies fall almost in a straight line. Average sac-cade lengths show similar developmental patternsbetween the two studies. These results, along withthe good performance in reading comprehensionquestions, suggest that our samples—both in termsof participants and texts—are representative ofrespective age levels within the languages.

Development of Reading Eye Movements

We hypothesized three possible developmentalpaths for reading eye movements. First, eye move-ments may become more efficient because ofimprovements in overall speed of processing or ingeneral reading-related processes. Correspondingly,there would be no differences in eye movements inreading different orthographies. Alternatively, themechanism that controls reading eye movementsmay be orthography specific, particularly for begin-ning readers who are yet to be freed from effortfulword recognition processes. If this is the case, we

should observe distinct eye movement patternsamong young readers of English and Chinese.Finally, different components of the eye movementmechanism may follow distinct developmentalpatterns, some of which are language-specific andothers language-universal.

Results from the present study provide unequiv-ocal support for the last view. Clearly, different eyemovement parameters follow different developmen-tal trajectories; three patterns emerge from the data:

1. Various saccade-related measures continue todevelop throughout elementary school, andshow considerable differences between Eng-lish- and Chinese-speaking readers. Consis-tent with previous research (Fischer et al.,1997; Klein & Foerster, 2001), this suggeststhat voluntary control in the where system islinked to the development of cognitiveprocesses—in this case increasingly proficientreading processes. Consequently, differencesin saccade patterns reflect different perceptualand cognitive demands Chinese and Englishreading imposes on the reader. This will befurther elaborated in the following section.

2. Beginning readers make higher proportionsof long fixations; this pattern is invariantacross the two languages and orthographies,suggesting a general developmental trend.Although the present study cannot ruleout the possibility that this is due to matura-tion, a recent theory sheds light on thenature of the phenomenon. In their competi-tion–inhibition theory, Yang and colleagues(Yang, 2006; Yang & McConkie, 2001) pro-posed that fixation duration during readingis determined by the competition betweentwo brain centers (the fixate center and themove center). In most cases, fixation durationis determined autonomously by the interac-tion between these two centers without inter-ference from higher order processes. Butwhen necessary, for example, when compre-hension difficulties occur, the cognitive pro-cess may intervene by changing theinteraction between the two centers and thusextend the current fixation. Assuming begin-ning readers encounter more difficulties inreading, the competition–inhibition theorypredicts more cognitive intervention andthus a higher proportion of long fixations. Tothe extent that the above assumption is truein both English and Chinese reading, thetheory also nicely accounts for the lack of


language differences in the developmentalpattern of fixation durations.

3. Distributions of fixation duration reveal that,with the exception of long fixations discussedearlier, the basic mechanism for fixation con-trol appears to be stable across age and lan-guage. This finding corroborates observationsfrom fixation duration distributions inMcConkie et al. (1991). It is also consistentwith Fischer et al.’s (1997) conclusion that theneurological mechanism that controls fixationduration (the when system) develops earlyand is mature by school age. Thus, evidencefrom this and other studies suggests that: (a)the development of fixation duration controlis primarily a maturational process that isaccomplished by the time children learn toread, and (b) experiential factors (such asreading experience or orthography) have lim-ited influence on this process.

The above conclusion may appear to be in con-flict with the notion that fixation duration reflectsmoment-by-moment cognitive processes in read-ing (Rayner, 1998; Reichle, Pollatsek, Fisher, &Rayner, 1998; Reichle, Rayner, & Pollatsek, 2003).Two clarifications help to resolve the paradox.First, our findings on the rate of refixations andregressions support the notion that cognitiveprocesses affect eye movement planning in realtime. The data also suggest that the cognitiveinfluence may be limited to some fixations,leaving other fixations controlled by autonomousoculomotor processes (cf. Yang & McConkie, 2001;Yang, 2006).

Toward a Theory of Reading Eye MovementDevelopment

The existence of three distinct development tra-jectories in reading eye movements calls for a the-ory of the development of reading eye movementcontrol. Findings from the present study rule outthe simplistic notion that all aspects of reading eyemovements continue to improve with age and read-ing experience. In the remainder of this article, wewill provide an initial sketch of a model of readingeye movement development that is consistent withobservations from this and prior studies (e.g.,McConkie et al., 1991).

We speculate that a beginning reader isequipped with the basic oculomotor abilities tomaintain fixations and to plan and execute saccadesto selected targets. This seems to be a safe assump-

tion given that a typical 6-year-old should alreadyhave made well over 120 million eye movements(assuming 2 eye movements per second and anaverage of 8 hr of wakeful time per day). This isalso consistent with the finding of McConkie et al.(1991) that first-grade students follow the sameoculomotor strategies in targeting words as adultreaders. The present study and Fischer et al. (1997)provide evidence that basic fixation duration con-trol is mature by around the time children start toread, if not earlier.

In addition, children have some cognitive controlover these oculomotor functions, although theextent to which cognitive processes can overwritethe default oculomotor programs vary with age, theoculomotor function to be controlled, and the nat-ure of the cognitive processes involved. Withregard to age, Fischer et al. (1997) found that adultsare faster and better at inhibiting reflexive oculomo-tor programming; this should extend naturally toreading. Furthermore, both Fischer et al. and ourdata suggest that the when system that controls fix-ation duration is less susceptible to cognitive con-trol than the where system that is responsible fordetermining saccade targets. Thus, larger develop-mental and orthographic differences are observedin saccade-related measures than in fixation dura-tion. This distinction is important for understand-ing reading development and the underlyingneuropsychological mechanisms.

Finally, the driving force behind most develop-mental changes in reading eye movements is likelyto be changes in reading processes that reflectimproved reading proficiency. Because processessuch as word recognition and text comprehensionare highly efficient and automated among skilledreaders, there is less need for higher level processesto override default oculomotor programs, whichprovides an explanation for the remarkable similar-ities in the eye movements of skilled English andChinese readers.

Our results support the strategy of looking atyoung readers to understand the effects of orthog-raphy on the cognitive and perceptual mechanismsunderlying reading. Reading is a complex processthat involves the interaction of a host of perceptual,linguistic, and cognitive factors. Anyone who hasread this article to the end surely qualifies as anexpert in the coordination of these processes. Thisvery expertise makes disentangling the componentsof reading difficult. Children, on the other hand,are likely to be much more dependent on andaffected by the vagaries of the orthography they arelearning to read.


References

Andrews, S., Miller, B., & Rayner, K. (2004). Eye move-ments and morphological segmentation of compoundwords: There is a mouse in mousetrap. European Journalof Cognitive Psychology, 16, 285–311.

Bertram, R., & Hyona, J. (2003). The length of a complexword modifies the role of morphological structure: Evi-dence from eye movements when reading short andlong Finnish compounds. Journal of Memory and Lan-guage, 48, 615–634.

Buswell, G. T. (1922). Fundamental reading habits: A studyof their development (Vol. 21). Chicago: University ofChicago Press.

Carver, R. P. (1990). Reading rate: A review of research andtheory. San Diego, CA: Academic Press.

Chen, H.-C., Song, H., Lau, W. Y., Wong, K. F. E., &Tang, S. L. (2003). Developmental characteristics of eyemovements in reading chinese. In C. McBride-Chang &H.-C. Chen (Eds.), Reading development in Chinesechildren (pp. 157–170). Westport, CT: Praeger.

DeFrancis, J. (1989). Visible speech: The diverse oneness ofwriting systems. Honolulu: University of Hawaii.

Everson, M. E. (1986). The effect of word-unit spacing uponthe reading strategies of native and non-native readers ofChinese: An eye-tracking study. Unpublished Ph.D. dis-sertation, Ohio State University, Columbus.

Feng, G. (2006a). Eye movements as time-seriesrandom variables: A stochastic model of eye move-ment control in reading. Cognitive Systems Research, 7,70–95.

Feng, G. (2006b). Eye movements in Chinese reading. In P.Li, L. Tan, E. Bates, & O. J. L. Tzeng (Eds.), Handbookfor East Asian psycholinguistics, (Vol. 1, pp. 187–194).Cambridge, UK: Cambridge University Press.

Feng, G. (2008). Orthography and eye movements: Theparaorthographic linkage hypothesis. In K. Rayner, D.Shen, X. Bai, & G. Yan (Eds.), Cognitive and culturalinfluences on eye movements (pp. 395–420). Tianjin,China: Tianjin People’s Press ⁄ Psychology Press.

Feng, G., Miller, K., Shu, H., & Zhang, H. (2001). Rowedto recovery: The use of phonological and orthographicinformation in reading Chinese and English. Journal ofExperimental Psychology: Learning, Memory, & Cognition,27, 1079–1100.

Findlay, J. M., & Walker, R. (1999). A model of sac-cade generation based on parallel processing andcompetitive inhibition. Behavioral & Brain Sciences, 22,661–721.

Fischer, B., Biscaldi, M., & Gezeck, S. (1997). On thedevelopment of voluntary and reflexive components inhuman saccade generation. Brain Research, 754, 285–297.

Gray, W. S. (1956). The teaching of reading and writing; aninternational survey. Paris: UNESCO.

Hoosain, R. (1991). Psycholinguistic implications for linguisticrelativity: A case study of Chinese. Mahwah, NJ: Erlbaum.

Hyona, J., Bertram, R., & Pollatsek, A. (2004). Are longcompound words identified serially via their constit-uents? Evidence from an eye-movement-contingent

display change study. Memory and Cognition, 32, 523–532.

Inhoff, A. W., & Liu, W. (1998). The perceptual span andoculomotor activity during the reading of Chinese sen-tences. Journal of Experimental Psychology: Human Percep-tion and Performance, 24, 20–34.

Inhoff, A. W., Radach, R., & Heller, D. (2000). Complexcompounds in German: Interword spaces facilitate seg-mentation but hinder assignment of meaning. Journal ofMemory and Language, 42, 23–50.

Judd, C. H., Gray, W. S., Gray, C. T., McLaughlin, K.,Schmitt, C., & Gilliland, A. R. (1918). Reading: Its natureand development. Chicago: University Chicago Press.

Kail, R. (1986). Sources of age differences in speed of pro-cessing. Child Development, 57, 969–987.

Klein, C., & Foerster, F. (2001). Development of prosac-cade and antisaccade task performance in participantsaged 6 to 26 years. Psychophysiology, 38, 179–189.

Kliegl, R. M., Olson, R. K., & Davidson, B. J. (1982).Regression analyses as a tool for studying reading pro-cesses: Comment on Just and Carpenter’s eye fixationtheory. Memory & Cognition, 10, 1287–1296.

Liu, Y., Tan, Q., & Shen, X. (1993). Xinxi Chuli YongXiandai Hanyu Fenci Guifan [Modern Chinese word seg-mentation rules for information processing purposes].Beijing, China: China Standard Press.

Liu, I. M., Yeh, J. S., Wang, L. H., & Chang, Y. K. (1974).Effects of arranging Chinese words as units on readingefficiency. Acta Psychologica Taiwanica, 16, 25–32.

McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D.(1988). Eye movement control during reading: I. Thelocation of initial eye fixations on words. VisionResearch, 28, IN1–IN2.

McConkie, G. W., & Rayner, K. (1975). The span of theeffective stimulus during a fixation in reading. Percep-tion & Psychophysics, 17, 578–586.

McConkie, G. W., Zola, D., Grimes, J., Kerr, P. W., Bryant,N. R., & Wolff, P. M. (1991). Children’s eye movementsduring reading. In J. F. Stein (Ed.), Vision and visual dys-lexia (pp. 251–262). London: Macmillan.

Miles, W. R., & Shen, E. (1925). Photographic recordingof eye movements in the reading of Chinese in verticaland horizontal axes: Method and preliminary results.Journal of Experimental Psychology, 8, 344–362.

Miller, K. F., Chen, S.-Y., & Zhang, S. (2007). Where thewords are: Judgments of words, phrases and syllables byspeakers of Chinese and English. Unpublished manuscript.

Morris, R. K., Rayner, K., & Pollatsek, A. (1990). Eyemovement guidance in reading: The role of parafovealletter and space information. Journal of ExperimentalPsychology: Human Perception and Performance, 16, 268–281.

Nagy, W. E., Anderson, R. C., & Herman, P. A. (1987).Learning word meanings from context during normalreading. American Educational Research Journal, 24, 237–270.

Peng, D. L., Orchard, L. N., & Stern, J. A. (1983). Evalua-tion of eye movement variables of Chinese and


American readers. Pavlovian Journal of Biological Science,18, 94–102.

Rayner, K. (1986). Eye movements and the perceptualspan in beginning and skilled readers. Journal of Experi-mental Child Psychology, 41, 211–236.

Rayner, K. (1998). Eye movements in reading and infor-mation processing: 20 years of research. PsychologicalBulletin, 124, 372–422.

Rayner, K., Sereno, S. C., & Raney, G. E. (1996).Eye movement control in reading: A comparisonof two types of models. Journal of ExperimentalPsychology: Human Perception and Performance, 22, 1188–1200.

Reichle, E. D., & Perfetti, C. A. (2003). Morphology inword identification: A word-experience model thataccounts for morpheme frequency effects. ScientificStudies of Reading, 7, 219–237.

Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K.(1998). Toward a model of eye movement control inreading. Psychological Review, 105, 125–157.

Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-ZReader model of eye-movement control in reading:Comparisons to other models. Behavioral and BrainSciences, 26, 445–476.

Seymour, P. H. K., Aro, M., & Erskine, J. M. (2003). Foun-dation literacy acquisition in European orthographies.British Journal of Psychology, 94, 143–174.

Shu, H., Chen, X., Anderson, R. C., Wu, N., & Xuan, Y.(2003). Properties of school Chinese: Implications forlearning to read. Child Development, 74, 27–47.

Stevenson, H. W., Lee, S.-y., Chen, C., Stigler, J. W., Hsu,C. C., & Kitamura, S. (1990). Contexts of achievement:A study of American, Chinese, and Japanese children.Monographs of the Society for Research in Child Develop-ment, 55.

Sun, F., & Feng, D. (1999). Eye movements in readingChinese and English text. In J. Wang, A. W. Inhoff, &H.-C. Chen (Eds.), Reading Chinese script: A cognitiveanalysis (pp. 189–206). Mahwah, NJ: Erlbaum.

Sun, F., Morita, M., & Stark, L. W. (1985). Com-parative patterns of reading eye movement inChinese and English. Perception & Psychophysics, 37,502–506.

Taylor, S. E. (1965). Eye movements in reading: Facts andfallacies. American Educational Research Journal, 2, 187–202.

Treiman, R., Weatherston, S., & Berch, D. (1994). The roleof letter names in children’s learning of phoneme-

grapheme relations. Applied Psycholinguistics, 15, 97–122.

Tsai, J. L., Lee, C. Y., Hung, D. L., & Tzeng, O. J. L. (2001,August). Eye movement guidance without word space: thepreferred landing location in reading Chinese text. Paperpresented at the European Conference on Eye Move-ments, Finland.

Tsai, C. H., & McConkie, G. W. (1995). The perceptual spanin reading Chinese text: A moving window study. Paperpresented at the International Conference on theCognitive Processing of Chinese and Other AsianLanguages, Hong Kong.

Tsai, J. L., & McConkie, G. W. (2003). Where do Chinesereaders send their eyes? In J. Hyona, R. Radach & H.Deubel (Eds.), The mind’s eyes: Cognitive and appliedaspects of eye movements (pp. 159–176). Amsterdam:North-Holland ⁄ Elsevier.

Tsai, J. L., Tzeng, O. J. L., Hung, D. L., & Yen, N. S.(2000). The perceptual span in reading Chinese passages: Amoving window study of eye movement contingent display.Paper presented at the annual meeting of the ChinesePsychology Association, Taipei, Taiwan.

Twain, M. (1997). The awful German language: Appendix Din A tramp abroad. New York: Penguin. (Original workpublished 1880)

Venezky, R. L. (1999). The American way of spelling: Thestructure and origins of American English orthography.New York: Guilford.

Wang, L. (2003). Hanyu Ci de Shehui Yuyanxue Yanjiu [Asociolinguistic study on Chinese words]. Beijing:Shangwu Press.

Yang, H. M. (1994). Word perception and eye movements inChinese reading. Unpublished doctoral dissertation,University of Illinois at Urbana-Champaign.

Yang, H.-M., & McConkie, G. W. (1994). Eye movementcontrol in Chinese reading. Bulletin of the NationalTaiNan Teacher’s College, 29, 193–229.

Yang, H.-M., & McConkie, G. W. (1999). ReadingChinese: Some basic eye-movement characteristics. In J.Wang, A. W. Inhoff, & H.-C. Chen (Eds.), ReadingChinese script: A cognitive analysis (pp. 207–222).Mahwah, NJ: Erlbaum.

Yang, S.-N., & McConkie, G. W. (2001). Eye movementsduring reading: a theory of saccade initiation times.Vision Research, 41, 3567–3585.

Yang, S.-N. (2006). An oculomotor-based model of eyemovements in reading: The competition ⁄ Interactionmodel. Cognitive Systems Research, 7, 56–69.


Appendix A. Reading Materials

Language Story No. of words (char) Word length Grade 3 Grade 5 Adults

English BOAT 60 5.35 X X X

MULE 222 5.65 X X X

BASKETBALL 87 5.40 X X

BEES 264 4.90 X X

GOODALL 385 5.65 X X

CELL 108 6.32 X X

MOZART 249 5.07 X X

NEWTON 429 5.96 X X

Chinese BOAT 61 (83) 1.62 X X X

MULE 196 (250) 1.56 X X X

BASKETBALL 68 (83) 1.48 X X

FISH 190 (234) 1.55 X X

MAO 304 (426) 1.54 X X

SWALLOW 129 (162) 1.67 X X

WU 248 (311) 1.57 X X

QIAN 349 (534) 1.70 X X

Note: Further information about the reading materials is provided below. All materials are available upon request.

Two stories were chosen to be comparable across languages. The ‘‘Power of Boats’’ (BOAT, see Appendix B) was an expository text on

the evolution of boat engines, translated from English to Chinese. ‘‘The Foolish Mule’’ (MULE) was an Aesop fable about a mule who

soaked a sack of cotton when crossing a stream in hope to lessen the weight on his back. We used two roughly parallel versions of the

same story in both cultures.

The rest of the stories were selected from extracurricular reading books published in China and the United States. They were represen-

tative of readings children would naturally encounter everyday. For the English materials, BASKETBALL was a short passage about

the origin of the basketball game. BEES and CELL introduced some interesting nature of honeybees and the history of the discovery of

cells, respectively. GOODALL, NEWTON, and MOZART were bibliographical stories about two scientists, Jane Goodall and Sir Isaac

Newton, and a musician, Wolfgang A. Mozart.

The Chinese materials were similar to the English ones in terms of contents and length. BASKETBALL was also about the history of

basketball, although the story differed in details from the English version. FISH and SWALLOW were expository pieces on deep sea

fish and the migration of swallows, respectively. There were also three bibliographical stories, as in the English stories. MAO was a

story about Yisheng Mao, an architect who is known for his design of many bridges in China. QIAN was about Xuesen Qian, a well-

known Chinese physicist. A story chosen to be parallel to MOZART, WU was a bibliography on Cheng’en Wu, a famous writer in

Ming Dynasty who wrote the novel XiYouJi, whose main character, the Money King, is loved by children in China.

Appendix B. A Sample Story: The Power of Boats

The English version

Different boats use different kinds of energy for power. In the past, boats usually used sails to harnessthe power of the wind. Other boats had oars and needed people for power. But many improvements weremade over time. The invention of the engine, for example, allowed boats to travel faster and farther. Nowsome boats use even better motors.

The Chinese version


orthography and the development of reading processes: an

Documents