UC Irvine UC Irvine Previously Published Works
Title Mathematics achievement of Chinese, Japanese, and American children: ten years later.
Permalink https://escholarship.org/uc/item/8rh630g0
Journal Science (New York, N.Y.), 259(5091)
ISSN 0036-8075
Authors Stevenson, HW Chen, C Lee, SY
Publication Date 1993
DOI 10.1126/science.8418494
License https://creativecommons.org/licenses/by/4.0/ 4.0
Peer reviewed
eScholarship.org Powered by the California Digital Library University of California ARTICLES
Zool. inst. Akad. nauk SSSR 131, 104 (1985). 1985), pp. 225-241; C. T. Keally, Curr. Res. exhibit evidence of entrance tunnels (57). 57. N. N. D1kov. Arkheotcgicheskie pamyatniki Kam Pleistocene 7, 143 (1990) 82. J . H. Greenberg, C. G. Turner II , S. L. Zegura, chatki. Chukotki I Verkhne1 Kotymy (Nauka, Mos 81. Some of the contrasts in the Beringian archaeo Curr. Anlhropol. 27, 477 (1986). row. 1977). logical record between the occupations associat· 83 Support for our research in the northern foothills of 58. A. V. Lozhkin and V. P. Parii, Opyt radiouglerod ed with the 12.000 to 11 ,000 yr B.P. interstad1al the Alaska Range was provided by the National nogo datirovamya verkhnechetvertichnykh otloz and those associated with the 11 .000 to 9.500 yr Geographic Society. the National Science Foun henii (Akademiya nauk SSSR. Magadan, 1985) B.P. stadiaJ may reflect chmahc differences. dahon, Alaska Division of Parks, the Unive
SCIENCE • VOL. 259 • I JANUARY 1993 53 Flg .1. (A) Fifth graders' performance on mathematics 20 test given at three testing periods for students from .,35 A ~ ~ !J (A), (0 ), Japan Taiwan and the United States (• ). ~ 15 Standard deviations for 1980: 6.45, 4.52, and 4.70, ~30 ~ .." respectively. The comparable values for 1984 were c =.. 10 5.82, 5.68, and 5.07 and for 1990were 5.34, 5.41, and : 25 0 5.21 . Sample sizes: 239, 240, and 244 (1980); 165, ::E 164, and 118 (1984); and 273, 242, and 241 (1990) . In 1980, the Japanese-American difference was 8.36 ± 1980 1984 1990 0.52; the Chinese-American difference, 5.83 ± 0.42; 10 20 30 40 50 and the Japanese-Chinese difference, 2.53 ± 0.51 . In 11 B Score 1984, the respective differences were 5.39 ± 0.65, Fig . 2. Eleventh grader's scores on the mathe 7.59 ± 0.64, and 2.19 ± 0.63; in 1990, 6.92 ± 0.47, f 10 0 matics test. Japan, heavy line; Taiwan, dashed 9.65 ± 0.48, and 2.73 ± 0.48. (B) Fifth graders' ~ line; United States, light line. Mean ±1 standard performance on the reading vocabulary test given at c 9 : deviation: 21.72 ± 6.59, 24.10 ± 9.47, and two testing periods. Standard deviations for 1980: ::E 13.39 ± 7.06, respectively. Sample sizes 1120, 8 3.39 (Japan), 1.97 (Taiwan), and 2.64 (United States); 1475, and 1197. for 1990: 2.39, 1.70, and 2.58, respectively. Sample sizes: 241, 241, and 246 (1980); 273, 240, and 242 1980 1990 (1990). In 1980, the Japanese-American difference was 0.77 ± 0.28; the Chinese-American difference, 1.57 ± 0.21 ; and the Japanese-Chinese difference, 2.34 ± 0.25. In 1990, the respective differences were 1.55 ± 0.22, 0.68 ± 0.20, and 0.87 ~0.5~ ± 0.18. 'E 0 CD 'C c over the past decade and to evaluate the We constructed all tests, interviews, and ~-0.5 status of American students as they progress questionnaires used in the study. The tests, c from elementary school to high school. based on detailed analyses of the textbooks CD ::E.. used in each city, were highly reliable (6) . · 1 ~------:!5:------1,.,_1 Outline of the Study By including only concepts and operations Grade common to all three locations, we reduced Our sample in 1980 included approximately the potential problem of differential expo Fig. 3. Mean standard scores on the mathemat 240 first graders and 240 fifth graders in sure, as discussed recently by Westbury (7). ics tests at grades 1, 5, and 11 for the longitu each city. Nearly half of the students who The same mathematics and reading tests dinal sample. Standardization procedures were conducted separately for each grade level. were first graders in 1980 attended fifth were given to the fifth graders in all three Sample sizes for Japan (6.) , Taiwan (0), and grade at the same schools in 1984, the time testing periods. United States (• ). respectively: 240, 241, and of our first follow-up. We also included a test of general infor 237 (1980); 165, 164, and 118 (1984); and 82, In 1990 and 1991, we attempted to mation. This test provided a measure of 156, and 198 (1990). The Japanese-American follow up the first graders in our original knowledge that children did not usually difference at first grade was 0.53 ± 0.09, and study (now in eleventh grade). We located learn in school but had acquired through the Chinese-American difference was 0.73 ± and obtained the cooperation of 212 stu their everyday experiences. The test for 0.09. At fifth grade the respective differences dents in Minneapolis, 169 in Taipei, and elementary school students included items were 0.84 ± 0.10 and 1.19 ± 0.10; at eleventh 93 in Sendai. Japanese eleventh graders such as "What are two things a plant needs grade, 0.80 ± 0.10 and 0.87 ± 0.11 . were reluctant to participate in the study in order to grow?" and "Why can't people after school hours, primarily because they live under water?" The eleventh-grade test children in 1990 had scores as low as those were preparing for the college entrance included more difficult items, such as "Why of the average American child. examinations they would take the following has it become possible to make smaller Reading ability also differed among stu year. To place the eleventh-grade follow-up computers in recent years?" and "What do dents in the three cities. In 1980, Chinese students in a broader context, we also we mean by inflation when we talk about a children obtained the highest average score studied cross-sectional samples totaling country's economy?" (8). on reading vocabulary, and Japanese chil nearly 4000 students in Minneapolis, The interviews and questionnaires cov dren received the lowest scores. In 1990, Taipei, and Sendai. These students attend ered parental attitudes and beliefs and many Japanese fifth graders had become the top ed 9, 18, and 8 different high schools, aspects of children's lives at home and after performers, and American students re respectively. A larger number of schools school. We included some common ques ceived the lowest scores (Fig. lB). was necessary to obtain a representative tions in the interviews and questionnaires Eleventh graders. Of the three groups, sample in Taipei because there are separate constructed for each grade level and for American students continued to receive the schools for boys and girls, and there are each aspect of the study. lowest scores in mathematics at eleventh both regular and vocational high schools. grade. Distributions of the scores of 3 792 New samples of approximately 240 fifth Academic Achievement eleventh graders in the cross-sectional sam graders in each city were also studied in ples (Fig. 2) reveal a wide disparity between 1990 (5). Fifth graders. The achievement gap in math the average performance of the American We interviewed over 85% of the moth ematics persisted throughout the decade and of the Chinese and Japanese students. ers of the elementary school students and of (Fig. IA). Japanese children were consis Only 14.5% of the Chinese and 8.0% of the longitudinal sample of eleventh graders tently superior to the American children, the Japanese students received scores below in each city. Over 73% of the fathers of the and the difference between the performance the average score of the American students. eleventh graders in the longitudinal sample of the Chinese and American children was The distribution of scores in Taipei was filled out a questionnaire. We also inter greater in 1990 than in 1980. The gap is bimodal. This was related to the type of viewed the mathematics teachers of the large: for example, only 4.1% of the Chi school the students attended. The average eleventh graders. nese children and 10.3% of the Japanese score of students enrolled in vocational
54 SCIENCE • VOL. 259 • l JANUARY 1993 cities. For example, at fifth grade the mean 60 ~ 0.5 score of the American top achievers was 1980 ~ 1.00 SD below the mean for the Asian top 60 1! achievers but only 0.04 SD higher than the 'O• 40 c mean of all of the Asian students. The same ! effect occurred at eleventh grade: the cop 20 :c -o. American students were L 13 SD below the ::::E mean of the cop Chinese and Japanese 01...-1-...... u.~- 5 11 students and only 0.3 7 SD above the mean 80 Grade of au Asian students. e 1984 Generally, boys and girls were equally .c• 60 Ag. 4. Mean standard scores for the test of capable in mathematics during elementary ~ general information at grades 1, 5, and 11 for school. This was true in all three cultures. 0 40 the longitudinal sample. Sample sizes for Ja At eleventh grade, however, scores of the c pan (&), Taiwan (0 ), and United States (• ). B 20 respectively: 240, 241, and 237 (1980); 161, boys in all three locations were significantly ... 161, and 119 (1984); and 82, 156, and 200 higher than those of the girls, and the rf gender differences were greater among Chi (1990). 80 nese and Japanese than among American students (11). high schools was 18.3; for students enrolled General information. The general infor in the regular academic high schools it was mation test offers an interesting contrast to 30. 7- Even the vocational high school stu the test of mathematics, where performance dents obtained a higher average score than is highly dependent on academic instruc the Minneapolis students' average of 13.4 tion. Asian superiority was not evident in points. the general information scores (Fig. 4). The relative status of students at elev Rather than diverge as grade level in enth grade can also be compared with what creased, as happened with the mathematics Ag. 5- Attitudes of Japanese (solid), Chinese existed at first and fifth grades. For this scores, the average scores of the American, (open), and American (hatched) mothers to ward children's academic performance as re purpose, we computed standard scores at Chinese, and Japanese students on the gen flected in ratings on a three-point rating scale. each grade level for students in the longi eral information test became increasingly Sample sizes for Japan, Taiwan, and United tudinal sample (9). The data (Fig. 3) offer similar. At first grade, the average scores of States, respectively: 233, 197, and 214 (1980); no evidence of improvement in the status of the American students exceeded those of 157, 104. and 117 (1984); and 258, 238, and the American students as they moved from their Chinese peers by 0. 72 ± 0.09 SD and 220 (1990). first through eleventh grade ( 10). of their Japanese peers by 0.43 ± 0.08 SD. If we analyze the data from all students By eleventh grade, the scores differed by tested at each grade level rather than from only 0.13 ± 0. 11 and 0.17 ± 0.13 SD, were asked about their children's academic only those in the smaller longitudinal sam respectively. Early American superiority achievement. Few Chinese and Japanese ple, we find chat the achievement gap in was also found in the performance of nearly mothers, but over 40% of the American mathematics increased between first and 900 kindergarten children in the three cit mothers, expressed high degrees of satisfac eleventh grades. At first grade, the average ies (12). Scores of the American kindergar tion with their children's academic perfor Chinese and Japanese scores differed from ten children exceeded those of the Chinese mance. One reason for their uncritical at the American scores by 0. 72 ± 0.09 and children by 1.33 ± 0.07 SD and of the titude may have been their lack of informa 0.53 ± 0.09 standard deviation (SD) com Japanese children by 0.69 ± 0.06 SD. tion about the relative status of American puted in the manner indicated above. By Thus, throughout their schooling Ameri children compared to their peers in other eleventh grade the corresponding differ can students proved to be as capable as or industrialized countries. Since 1980, how ences were L.18 ± 0.04 and 0.92 ± 0.03 even more capable than the Asian students ever, the media has disseminated many SD. The increased divergence of scores, when they were tested with items not based reports of the weaknesses of American ed 0.45 ± 0.10 for the Chinese-American on the school curriculum. ucational systems and of American students comparison and 0.39 ± 0.09 for the Japa We attribute the early superiority of the in international comparisons. We can ask, nese-American comparison, was statistical American children to the greater cognitive therefore, whether the publication of these ly significant. stimulation provided by their parents, who critical reports may have led to diminished Another question sometimes raised is indicated that they read more frequently to satisfaction by American mothers. It is whether average scores offer the most ap their young children, took them on more evident (Fig. 5) that it did not. If anything, propriate measure for comparing children's excursions, and accompanied them to more somewhat more American mothers said levels of achievement in different countries. cultural events than did the Chinese or they were "very satisfied" with their chil For example, might not the scores of high Japanese parents (13). As American chil dren's performance in 1990 compared with achievers, say the top 10% of students in dren grow older, parents appear to be less 1980. each location, be comparable? We sought likely to provide the kinds of enriched The continued satisfaction of American to answer this question by comparing the out-of-school experiences that they did be mothers is surprising because they seemed scores of the top 10% of American fifth and fore the children entered first grade. to be aware of the country's low status in eleventh graders with those of the top 10% comparative studies. For example, we told of Chinese and Japanese students. Parents' Satisfaction the mothers of the eleventh graders, "Sev The performance of the top Minneapolis eral recent studies have compared students' students was more similar to that of the One of the most dismaying findings in our school achievement in different countries. average Taipei and Sendai students than it 1980 study was the high level of satisfaction One recent study compared the perfor was to that of the top students in chose expressed by American mothers when they mance in math of high school students from
SCIENCE • VOL. 259 • I JANUARY 1993 55 i hi diidMMG'Fi Iii FT a: raauam11:rn1wo
5 100 Mother :? Q) 80 5 0 E 60 0
cGI :.e
Fig. 7. Percent of Japanese (solid), Chinese (open), and American (hatched) mothers who believed their children's schools were doing a "good" or "excellent" job. Sample sizes for Japan, Taiwan, and United States, respectively: Japan Taiwan U.S. 234, 227, and 216 (1980); 159, 105, and 117 (1984); and 261, 237, and 220 (1990). Fig. 8. Scores mothers would expect their chil 5 dren to obtain (solid) and scores with which the Student mothers would be satisfied (hatched) on a about their children's level of achievement. hypothetical mathematics test. Sample sizes for One-third of the American mothers said Japan, Taiwan, and United States, respectively: they were "very satisfied" with their child's 256, 235, and 207 (grade 5) and 157, 206, and academic achievement, a strong contrast 191 (grade11). with the 10% of the Chinese and 2% of the Japanese mothers who said they were "very ~ .nAcademicl .MathematicsDI satisfied." Fathers expressed similar atti were doing·as well in school as their parents achievement achievement tudes: the percentages of fathers of eleventh and teachers wanted. Asian students were Fig. 6. Ratings made by parents and eleventh graders who were "very satisfied" were 37% less sure ( 14). grade students of the students' academic (Minneapolis), 8% {Taipei), and 6% (Sen Mothers' attitudes about the academic achievement and mathematics achievement on dai). curricula were equally stable over the dec a seven-point rating scale ranging from "much American students, as well as their par ade. Despite the publicity about the need below average" (1) to "much above average" ents, had a high regard for their own aca (7). Sample sizes for Japan (solid), Taiwan for reform, for upgrading standards, and for (open), and United States (hatched), respec demic accomplishments. For example, we assigning more homework, mothers of fifth tively: 157, 215, and 196 (mother); 131, 193, asked the eleventh graders and their moth graders were no more enthusiastic about and 155 (father); and 149, 225, and 200 (stu ers and fathers to rate the students' academ instituting changes in 1990 than they had dent). ic achievement and achievement in math been in 1980 or 1984. Over 80% of the ematics compared with corresponding American mothers in all three periods achievements of other students of the same thought the level of difficulty of the curric eight industrialized countries. Among the age. For all three sets of ratings (Fig. 6), the ulum was "just right." In 1990, less than countries, where do you think (American, average ratings made by the American stu 10% thought it was either too hard or too Taiwan, Japanese) high school students dents and their parents were above those easy. Very few Japanese or Chinese mothers ranked in math?" American mothers esti made by their Chinese and Japanese coun ever thought the curriculum was too easy, mated that American students' scores terparts and were consistently higher than but from 10 to 30%, depending on the year would fall between sixth and seventh place. the score of "4,'' which was defined as that and location, thought it was too difficult. Chinese and Japanese mothers were aware of average students. American mothers expressed little en of their students' high status: they estimated We also asked the parents about how thusiasm about increasing the amount of that students in Taiwan and Japan would be good a job they thought their children's homework. The great majority said the between second and third place. In other schools were doing in educating their child. amount was "just right." A little more than words, Americans appeared to be aware The American mothers of fifth graders were a quarter of the American mothers thought that American education is in trouble but consistently more positive than the Chi their fifth graders had too little homework, did not ascribe the phenomenon to their nese and Japanese mothers (Fig. 7). The both in 1980 (27%) and in 1990 (28%). own children. positive attitude persisted to eleventh Nearly a third of the Japanese mothers and One reason American parents may ex grade: 79% of the American mothers, but 10% of the Chinese mothers expressed this press such satisfaction is that they seldom only 44% of the Chinese and 48% of the attitude. These attitudes existed even receive clear, explicit information about Japanese mothers, rated their children's though students in Taipei and Sendai spend their children's standing in academic sub schools as doing a "good" or "excellent" much more time on homework than the jects. In the United States, most elementa job. American fathers were also very posi students in Minneapolis (I 5). ry school teachers convey their evaluations tive: 77% said their eleventh grader's In short, Minneapolis parents continued of children through general phrases, such as school was doing a "good" or "excellent" to be as satisfied in 1990 as they had been in "satisfactory" or "needs improvement" or job, but Chinese and Japanese fathers were 1980 with their children's academic through drawings of smiling or frowning no more positive than the mothers: 51% of achievement, the quality of education pro faces. As students progress through school, the Chinese and 40% of the Japanese fa vided by their children's schools, the cur however, the grading system becomes more thers gave their children's schools the top ricula, and the amount of homework as informative and parents should have a ratings. signed. In view of the persistence of such clearer idea of how their children are per The children assimilated these positive positive attitudes and of the continued poor forming. Even so, parents of the American attitudes. As early as fifth grade, American performance of American students, one eleventh graders remained very positive students expressed confidence that they must wonder about the degree of popular
56 SCIENCE • VOL. 259 • l JANUARY 1993 iJhi WU ; Hfflriiffl& ¥Hi W i&Wiiili&i.WEUI itiii ARTICLES
support that exists generally in the United ligence, home environment, and studying frequent indications of psychological distur States for extensive changes in elementary hard. Which do you think is the most bance, it would be reasonable co argue that and secondary education. important factor?" More Chinese and Jap high standards for academic achievement Many American parents do not have anese than American students thought may occur at unacceptable costs. Rather, it high standards for their children's academic studying hard was the most important factor was the American students who were more achievement. For example, we told the (59 and 72% versus 27%, respectively). In likely to express indications of distress. We mothers, "Let's say there is a math test in contrast, more than twice as many Ameri believe this occurred because American which there are 100 points. The average can as Chinese or Japanese students chose students do not have a clear idea about the score is 70. What score do you think your "a good teacher" (54% versus 18 and 14%, importance they should place on education. child would get?" We then asked, "What respectively). Chinese and Japanese students are expected score would you be satisfied with?" Asian We asked the teachers of the eleventh to devote themselves primarily to their mothers were less easily satisfied than the graders the same question. Among Japanese studies. American students, in contrast, are American mothers. The tendency in all teachers, 93% selected "studying hard." faced with many opposing demands. For three locations, especially among mothers Only 26% of the American teachers chose example, more than three times as many of fifth graders, was to expect that their this alternative. In contrast, the first choice American as Asian students have after child would receive an above-average score of 41 % of the American teachers but of school jobs (74% versus 21%) and more (Fig. 8). American mothers tended to give only 7% of the Japanese teachers was innate than twice as many have "dates" (over 85% the highest estimates and Japanese moth intelligence. Chinese teachers' choices fell versus 37%). The motivation for economic ers, the lowest. In responding to the second between those of the American and Japa independence and broad social experience, question, American mothers said they nese teachers. as well as the desire to engage in sports and would be satisfied with the score they ex In another set of questions, we asked the to assist with family chores, make it difficult pected their child would receive (16). Chi students how strongly they believed in such for American high school students to de nese and Japanese mothers required a score statements as "Everyone in my class has vote themselves wholeheartedly to their higher than the expected score to be satis about the same natural ability in math." studies. fied. American students disagreed to a signifi We found similar results when we asked cantly greater degree than did Chinese or Implications the fathers and students these questions and Japanese students (17). When mothers when we asked about reading as well as were asked this type of question, American Unless radical reforms can be instituted in mathematics. Americans consistently indi mothers strongly disagreed; Chinese and American educational systems, it seems cated that the score with which they would Japanese mothers were more neutral (18). unlikely that American students will lead be satisfied was about the same as or lower the world in mathematics by the end of this than the score they expected their children Psychological Adjustment century, one of the well-publicized goals or they would receive. adopted by the governors and President Critics of the academic success of Chinese Bush in 1990. American schools may cur Ablllty versus Effort and Japanese students often suggest that rently be fulfilling the roles expected of their high levels of performance come at them by the American public, but its ex In earlier reports, we presented evidence of great psychological cost. This criticism is pectations prove to be insufficient when cultural differences in the relative emphasis based, we suspect, on informal reports and judged by international standards-Ameri given to effort and ability in accounting for old data; we know of no recent comparative can students received significantly lower the academic achievement of elementary studies supporting this belief. We sought scores on a curriculum-based mathematics school children. Chinese and Japanese fifth such information in our study by asking the test than their Chinese and Japanese peers, graders and their mothers, following long eleventh graders to tell us about how fre there was no significant improvement in held tenets of Asian philosophy, stressed quently within the past month they had the scores of fifth graders on the same test the importance of hard work as the route to experienced feelings of stress, depression, given three times over a period of 10 years, success. American mothers, to a greater aggression, and somatic complaints of pos and the status of American students relative degree than Chinese and Japanese mothers, sible psychological origin, such as being to their Chinese and Japanese peers de emphasized the importance of innate abili unable to sleep. We also asked about the clined between the first and eleventh ty. We wondered whether this characteris frequency with which they felt nervous grades. tic would remain applicable when students when they took tests or when the teacher These differences cannot be attributed were in high school. Years of trying hard handed tests back. to differential sampling. Enrollment in and not succeeding or of finding that high Japanese, not American students, re school is nearly universal among first and ability cannot compensate for failure to ported the lowest frequencies of occurrence fifth graders in all three locations, and the study might lead eleventh graders and their of all these characteristics. American stu percentage of adolescents enrolled in high parents to modify their earlier beliefs. This dents reported the most frequent feelings of school is similar if vocational high schools did not prove to be the case. Eleventh stress, academic anxiety, and aggression. are included, as was the case in our study. graders in Japan and Taiwan remained The most common source of stress was Other factors were also carefully controlled, strong adherents to the belief that hard school. It was mentioned by 70% of the such as the time in the school year the data work was of primary importance, and their American students-more than three times were collected, testing conditions, and the American peers retained their strong belief as often as four other major sources (peers, relevance of the test items. in the importance of ability as a modifier of family, sports, and jobs). Chinese students We conclude that the achievement gap the effects of effort. expressed somewhat more frequent feelings is real, that it is persistent, and that it is Several types of evidence support this of depression and somatic problems (19). unlikely to diminish until, among other conclusion. For example, we told the elev These data do not support the Western things, there are marked changes in the enth graders, "Here are some factors that stereotype of Asian students as tense young attitudes and beliefs of American parents may influence students' performance in persons driven by relentless pressures for and students about education. American mathematics: a good teacher, innate intel- academic excellence. Had they exhibited parents appeared to be no more likely in
SCIENCE • VOL. 259 • 1 JANUARY 1993 57 EdPRiHi R ft•' It i, • UH j I I! Id J 'IBfM 3 ildhi liiUTI W&iiEfi&& r•w•u rmaaa
1990 and 1991 than they were in 1980 to ed. Reliabilities of the mathematics tests were numbers of girls were 526, 851, and 480. high: Cronbach alphas, a measure of test reliabil 12. H. W. Stevenson, S.-Y. Lee, T. Graham, in Hand believe that there is an urgent need for ity defined by psychological testing theory, book of Research on Early Education, 8. Spodek, educational reform. They did not seem to ranged from 0.86 to 0.95 for the elementary Ed. (Macmillan, New York, In press). be incensed by the low levels of perfor school test and from 0.91 to 0.95 for the high 13. H. W. Stevenson et al., Monogr. Soc. Res. Child mance by American students. Rather, they school test. The reading vocabulary test was Dev. Ser. No. 221 (1990), vol. 55. highly reliable in Minneapolis and Sendai but 14. The mean ratings on a seven-point scale for appeared to be pleased with their children's lower in Taipei. Cronbach alphas were, respec responses referring to parents and teachers were, academic achievement, to be satisfied with tively, 0.80, 0.88, and 0.61. respectively, 3.92 and 3.77 (Japanese), 4.20 and the job their children's schools were doing, 7. I. Westbury, Educ. Res. 22. 18 (1992). 4.26 (Chinese), and 5.27 and 5.00 (American), where 7 indicates "very true" and 4 indicates and to believe that children's innate abili 8. The general information test for the elementary school students contained 26 items; the eleventh "somewhat true." Sample sizes for Japan, Tai ties guide their course of progress through grade test contained 12 items. At first and filth wan, and the United States, respectively: 273, school. Attitudes and beliefs are difficult to grades. students proceeded through the test until 242, and 241 . The Japanese-American differ change. But the likelihood of improving they missed four successive items. Eleventh ences were 1.35 ± 0.13 (parents) and 1.23 ± graders responded to as many of the questions 0.12 (teachers); the Chinese-American differ the nation's competitive position through as possible within a 12-minute lime limit. The ences were 1.07 ± 0.14 (parents) and 0.74 ± better education depends, at least in part, reliability of the test ranged from O. 79 to 0.91 for 0.13 (teachers); and the Japanese-Chinese differ on changing such optimistic but ultimately first and filth graders; at eleventh grade it was ences were 0.28 ± 0.14 (parents) and 0.49 ± 0.67 (Sendai), 0.80 (Taipei), and 0.82 (Minneap 0.12 (teachers). self-defeating views. olis). 15. C. Chen and H. W. Stevenson, Child Dev. 60, 551 (1989). 9. We determined standard scores separately for REFERENCES AND NOTES 16. The differences between the satisfactory and ex each grade by combining data for students in all pected scores for the United States. Taiwan, and three locations in a single distribution. We then Japan were, respectively, - 0.79 ± 0.69, 8.51 ± 1. H. W. Stevenson, S.-Y. Lee, J. W. Stigler, Science computed the standardized scores at each grade 0.60, and 11.54 ± 0.73 (grade 5); and 1.49 ± level using the mean and standard deviation of 231, 693 (1986). 1.00, 10.47 ± 0.87, and 10.08 ± 0.94(grade11). 2. Kids Count Data Book (Center for the Study of the weighted sample to yield equal proportions of 17. On a seven-point scale, the mean ratings for the students from each of the three locations. Social Policy, Washington, DC, 1989). Japanese, Chinese. and American students were 3. The State of Mathematics Achievement (National 10. Longitudinal comparisons across the 10-year 3.80, 3.84, and 2.86. On this scale, 1 indicated Center for Educational Statistics, U.S. Department span assume that the eleventh-grade sample of "strongly disagree" and 4 indicated "neither of Education, Washington, DC. 1991). students was not biased. This assumption proved agree nor disagree." Sample sizes for Japan, 4. H. W. Stevenson and J. W. Stigler, The Leaming to be valid for the Chinese and Japanese stu Taiwan, and the United States, respectively: 1060, Gap (Summit. New York, 1992). dents. Comparisons of the first-grade scores in 1472, and 1044. The Japanese-American differ 5. Consent was obtained from the American chil mathematics, reading, and a cognitive test com ence was 0.95 ± 0.07; the Chinese-American dren's parents for their participation. In Japan and posed of ten subtests revealed no significant difference was 0.98 ± 0.06; and the Japanese Taiwan such rights of consent are vested in differences between eleventh graders included in Chinese difference was 0.04 ± 0.06. school authorities and teachers. The students and the longitudinal sample and those who were not. 18. On a seven-point scale, the mean ratings for the their parents were told that if there was anything For the American students, the average first Japanese, Chinese, and American mothers were they felt uncomfortable about answering, they grade scores of the 29 students we were unable 3.50, 2.97, and 1.74. Sample sizes for Japan, could go on to the next question. The examiners, to include in the follow-up sample were signifi Talwan, and the United States. respectively: 157, all native speakers and residents of each city, cantly lower on all three tests than were those of 215, and 194. The Japanese-American difference explained that the session could be terminated if the 212 follow-up students. On the mathematics was 1.76 ± 0.13; the Chinese-American differ the respondent wished. test, for example, the respective scores were encewas 1.23 ± 0.14; and the Japanese-Chinese 6. The mathematics test given to first and fifth grad 15.07 ± 1.04 and 17.30 ± 0.36; for the reading difference was 0.53 ± 0.16. ers had 54 items arranged in order of difficulty. test, the respective scores were 5.72 ;:t 1.29 and 19. These results are discussed 1n greater detail in o . Some items required only computation; others 10.37 ± 0.67. We conclude that the achievement Crystal et al., in preparation. required application of mathematical principles to scores of the American eleventh graders would 20. Our study was undertaken with S. Kitamura and S. word problems. We tested all children individual have been even lower if the 29 additional Ameri· Kimura of Tohoku Fukushi University in Sendai, ly. Any child who failed to answer four successive can students had been included. Japan, and C.-C. Hsu and H.-J. Ko of National items correctly was stopped. The eleventh-grade 11. The mean difference between the scores of the Cheng Kung University, Taiwan. The research mathematics test contained 46 items covering a eleventh-grade boys and girls In Minneapolis was was supported by a grant from the National broad range of problems. A 40-minute time limit 1.36 ± 0.45; in Taipei, 4.19 ± 0.49; and in Sendai, Science Foundation. Collection of the Taiwan data was imposed. Questions on both tests were open 5.43 ± 0.34. The respective numbers of boys in was supported by grants from the National Sci ended, and no multiple-choice items were includ- the three cities were 478, 617, and 639; the ence Council.
58 SCIENCE • VOL. 259 • l JANUARY 1993