MARCIA C. LINN CATHERINE LEWIS INEKO TSUCHIDA NANCY BUTLER SONGER

Educational Research 29 2000, No. 3, 4-14

Between 4th and 8th grade, American students fall behind on international norms, whereas Japanese students continue to perform well. This paper brings diverse perspectives to bear on Japanese late-elementary science education, in order to elucidate its instructional features and the broader educational system features that enable deep, coherent instruction.

Why do American students fall behind Japanese students in science performance after fourth grade? According to the Third International Mathematics and Science Study (TIMSS), Japanese and U.S. students perform similarly in science at fourth grade; but by seventh grade Japanese students remain near the top of all countries studied, while American students score at about the middle of all countries studied, and significantly lower than Japanese students: 23 points lower-nearly the equivalent of one U.S. grade level (Schmidt, McKnight, & Raizen, 1996; Schmidt, Raizen, Britton, Bianchi, & Wolfe, 1997).

To explore this divergence in science performance, we observed, recorded, and analyzed science lessons in 10 Japanese classrooms, and we reviewed a large body of evidence on Japanese instructional context, curriculum, and policy. We integrate information from these sources to identify some of the possible connections between science instruction and broader features of the educational system. Our intent is (a) to provide an account of Japanese science education that points out many differences in context between the U.S. and Japan, and thereby discourages simplistic inferences about Japanese science achievement; (b) to provide examples of the synergies between science instruction and broader educational system features in Japan; and (c) to explore the implications of such synergies for U.S. efforts to improve science education.

In this paper, we first describe the specific science activity structures we observed. Next we review research on Japanese educational system features. "Science activity structures" are what students do, individually or in groups, as they learn science-for example, experiments, data sharing, discussion of findings, assessments, etc. We sought recurring activity structures in the observed lessons, and tried to describe how these structures contribute to science learning.

We visited a convenience sample of seven schools in the Tokyo region that were conducting open research lessons (in which Japanese teachers observe each other; see below) or regular lessons in science and computers during the time we visited Japan. This paper excludes two computer lessons, and focuses on 10 science lessons taught in five schools. We visited ordinary local public schools and national public schools (selective-admission schools that provide leadership in science, as well as other subjects). We observed a sixth-grade lesson on "Air and How Things Bum" at one national elementary school and two ordinary elementary schools. We observed a sixth-grade lesson on "Electromagnets" and a fifth-grade lesson on "How Matter Dissolves" at national elementary schools. We observed fifth-grade lessons on "The Function of Levers," and "The Way Nature Works Lately-the study of Hydrangeas," at national elementary schools. We observed fourth-grade lessons on "Balance and the Weight of Things" and "The Body at Work" at national elementary schools. And, we observed a fifth-grade lesson on the "Development of Animals" at an ordinary elementary school. At each school we visited, we interviewed the classroom teachers we observed; in many cases we interviewed principals and other teachers as well.

Because our observations include research lessons at several elementary schools attached to national universities, the Japanese science lessons that we observed were probably closer to "leading" practice than to the representative teaching TIMSS is designed to capture (Azuma & Walberg, 1985; Tsuchida, 1993). Most of the teachers we observed specialized in science (i.e., they considered science their specialty even though they might teach all subjects) and had more than 5 years of teaching experience.

MARCIA C. LINN is a professor at the Graduate School of Education, University of California, Berkeley. Her areas of specialization include education technology, gender equity, and science education.

In each classroom that we visited, we gathered a range of information. Every teacher agreed to be videotaped; we used one or two video cameras. In every classroom, we audio-taped the classroom instruction and its English translation (performed by Tsuchida, a native Japanese speaker) and dictated information about classroom activities. In addition, one of the observers took time-stamped field notes continuously during the lessons while listening, to the English translation. Finally, we obtained copies of supporting materials (textbooks, teachers' manuals, handouts, student work) and took still photographs in each classroom to capture details of each lesson. Details of the observational method, activity structures in each lesson, and interview format are available from the first author.

For this paper the four observers discussed and synthesized an account of the science activity structures, The four investigators reviewed each other's interpretations and reached consensus. Disagreements were resolved by referring back to original data sources. The four participants included two bilingual researchers who have conducted extensive research in Japanese elementary schools for 10-20 years and two established science education researchers who have conducted extensive research and intervention in U.S. schools.

Results

We identified eight science activity structures and a rough order of occurrence in typical science lessons. Students moved easily through these structures; they -willingly reasoned about observed phenomena and transferred findings from one lesson to another. Table 1 describes the activity structures for the 10 lessons, with examples drawn from the lessons; it incorporates supplementary information on the lessons gained from teacher interviews and written materials.

Virtually every lesson began with an activity designed to connect the lesson to the interests of students or to elicit their current views based on their personal experience or on earlier lessons. During, this introduction, some teachers subtly clarified the scope of the science topic, redirecting students to promising connections. One teacher began by wondering out loud what the fish in the class aquarium ate.

Another brought a printed list of student reflections on dissolving (from an earlier lesson) and asked students to find similarities and elaborate on the various reflections.

Activity Structure 2: Elicit Student Ideas or Opinions

Teachers typically moved seamlessly from the first activity (connecting the science lesson to student interest and prior knowledge) to the second: broadly eliciting student ideas and opinions about what they know and what they need to find out. For example, in one class students were asked to visualize the activity of air in a closed jar as a candle burned. In another, the teacher asked students to think about what factors influence the force needed to lift a weight (a 5-kg bucket of sand) with a pole. Teachers elicited- students' ideas to determine questions students wanted to answer, and they encouraged students to frame concrete questions informed by past lesson . For example, in the lesson on levers, the teacher reminded students of their thinking that it was easier to pull a stake out of the ground with a long pole than a short pole, and connected the current question to an earlier experiment that the students had conducted. By incorporating students' own prior discoveries into class questions, the teacher guided students to focus on the relationship between the weight lifted and the distance from weight to fulcrum.

Activity Structure 3: Plan Investigations

Teachers assisted students in framing hypotheses, making predictions, designing experiments, and selecting methods questions elicited earlier from students. Often teachers described the equipment or materials available for the investigation. Students predicted outcomes and identified factors likely to influence the outcomes. They devised ways to assess their predictions.

Activity Structure 4: Conduct Investigations

Students conducted hands-on experiments only after considerable planning and preparation in the first three activity structures. They investigated hypotheses they had helped develop using methods they believed to be valid. For example, in one class students recorded their underarm body temperature, after which they ran up and down the stairs three times and recorded their body temperatures again. In another class students used poles, pails with 5 kg of sand, and clay; they created levers to hit the sand pails. They recorded the distances from pail and clay to fulcrum, along with their results. Small groups (usually four students; see discussion of the han below) gathered materials, made measurements, recorded results, cleaned up spills, and put away materials, with little teacher supervision.

Activity Structure 5: Exchange Information From Investigations

In every class, students recorded the results of their investigations and compared these across groups or individuals. In some cases, students questioned findings from other groups resolved inconsistencies. Often students posted their results on the blackboard in narrative, table, or outline formats. Sometimes each group reported findings verbally to the class, after which teachers synthesized their reports on the blackboard or in a subsequent handout. The results were made public so that variations in findings could be observed and discussed.

Activity Structure 6: Systematically Analyze or Organize Information

Teachers helped students find patterns in the class findings that had been posted using tables, visual highlighting, verbal summaries, and other strategies. In the body temperature and exercise experiment, students posted results on the blackboard and the teacher highlighted them, placing red magnets next to those results that showed decrease in body temperature. Teachers used a variety of strategies to highlight similarities, differences and other patterns in the data, and they also invited students to identify patterns.

Activity Structure 7: Reflect and Revisit Hypotheses or Predictions

To tie results back to the initial hypotheses and predictions, each teachers often asked students to write about their thinking. For example, after an experiment using oxygen detectors to study combustion, the teacher asked students to return to the question under investigation whether the oxygen in the jar had decreased or completely disappeared and to write their conclusions about it. Student answers such as these were studied by teachers in order to plan subsequent lessons; they also provided an assessment of student progress. In small group or whole-class discussion, students were asked to address the implications of findings. Often students were asked to explain their thinking to other students and to express their agreement or disagreement with others' ideas.

Activity Structure 8: Connect Results to Next Lesson

To carry forward interest and ideas, teachers asked students to identity conundrums, unanswered questions, and ancillary questions that emerged from their w dents suggested new questions that might be the focus subsequent lessons. For example, students who studied what happened to oxygen when a candle was burned in a closed container suggested studying changes in burning when oxygen is added to the container.

Activity Structures and Coherence

The activity structures we have just described seem to s port a coherent inquiry process by having students build on what they know or have investigated in class, design new investigations to fill in gaps in their understanding, reconcile conflicting findings, locate patterns in data, and reflect on progress. We observed teachers using a variety of classroom practices to implement each activity structure, including whole-class discussion, small-group work, and individual writing, experimentation or reflection. We rarely saw the teachers assign homework but frequently observed students spontaneously get out their science notebooks and record their ideas in words and pictures.

We were impressed by the lively discussions, frequent citing of related ideas from prior class work, and sustained interest in a cohesive account of the scientific topic under investigation. Students contributed to each other’s understanding by questioning and building on each other's ideas.

The sequence of activity structures we have synthesized above and described in Table I also fits well the 20 fourth-grade science lessons in Nagoya-area local public schools studied by Tsuchida (1993), and those described by Azuma and Walberg (1985), though it was not specifically designed to do so. The rubric also echoes the "script" for mathematics classes in Japan described by Stigler, Fernandez, & Yoshida (1996). Taken together, the activity structures comprise an inquiry process that may occur again and again as students learn science.

The investigators concurred that the topics, activities, and equipment observed in the 10 Japanese science lessons were not unique to Japan; they had much overlap with U.S. model science programs and with science lessons observed by the first author in several countries (Linn & Hsi, in press; Linn, Songer, & Eylon, 1996; Minstrell & Stimpton, 1996; Stake & Easley, 1978).

Despite many similarities to lessons observed in the U.S. and other countries, two aspects of the lessons were striking. First, each of the elements listed in Table 1 was implemented across all or nearl all of the 10 lessons, Because our convenience sample included leading practitioners, the consistency across lessons may have been greater than is generally the case across Japanese schools. Yet we were startled that, in five different schools in two different cities, across eight different science topics, we saw such consistency in science activity structures. For example, 9 of 10 sons began with a whole-class discussion in which teachers elicited students' prior knowledge of the topic to be studied, followed by students making observations), and sharing their data with classmates. Most lessons proceeded next to discussion of findings and revisiting hypotheses, though in some cases this would be done in a subsequent lesson. We doubted that, even if we selected practitioners with shared training or philosophy in the U.S., we would see such consistency in the approach to diverse science topics. It may be that, like the Japanese mathematics lessons (Stigler et al., 1996), Japanese elementary science lessons have a lesson structure that is widely shared across practitioners.

Evidence suggesting a widely shared set of science activity structures also came from teachers' discussions following research lessons. For example, the following conversation among teachers from different Japanese elementary schools followed a public research lesson at a national elementary school:

Visiting Teacher: Using information sheets to have students exchange findings is unique to this school, and different from the student group discussions or presentations that are more commonly used for student exchange of information. But how do you assist students who may not be able to comprehend or benefit from information presented in a written form?

Science Department Head: We think that exchanging written information promotes students' reflection more than verbal exchange does, since students easily forget what they hear. Written information exchange also encourages students to take initiative as they gather information. Students selectively collect from the sheets the information relevant to their interests.

The visiting teacher assumed a shared basis of practice across Japanese elementary schools: that science lessons routinely include student information exchange, and that most teachers accomplish this goal through student discussion or presentations, not written exchange. Perhaps because of the regular rotation of teachers among schools (Lewis, 1995), the study groups that involve teachers from many schools, and the relatively frequent opportunities to see others' lessons (Lewis & Tsuchida, 1997; Stigler & Eebert, 1997), Japanese teachers feel justified in assuming a shared set of science activity structures. Such conversations, which are a widespread feature of Japanese practice (Lewis & Tsuchida, 1997), may actually build consistent practice by giving teachers opportunities to make collective sense of educational goals and of the effective science activity structures for reaching them.

A second, striking feature of the observations was that the science activity structures seemed to be supported in a number of ways by broader features of the educational system. We review research on selected Japanese educational system features, followed by a discussion of the connections between system features and science activity structures.

Our review focuses on three system features germane to the observations: (a) emphasis on student character and attachment to school; (b) organization of schools for student and adult learning; and (c) the national course of study established through national guidelines, textbooks, and policy. To shed light on the science achievement differences that emerge between fourth and seventh grade, we focus mainly on elementary schooling.

Japanese elementary education has been described as "whole person education," with strong emphasis on children's social and ethical development and their capacity to function without teacher administered rewards or sanctions (Cummings, 1980; Lewis, 1995; Sato, 1991). Japanese elementary teachers seek to build these student qualities by (a) giving children considerable classroom authority (e.g., act as daily rotating class leaders) so that students will feel invested in school practices; (b) highlighting social and ethical qualities such as friendliness, responsibility, and persistence as central educational goals; and (c)organizing instruction in ways that meet children's needs for belonging and contributing, and thereby foster student’s attachment to school and their disposition to take on the school’s values as their own (Lewis, 1995)

Average class size is about 30, but the basic unit for many activities is the han: heterogeneous, family-like groups of about four children who share activities throughout the day, ranging from science experiments to art projects. Ethnographic and observational accounts suggest that, by the time of the fourth-grade TIMSS measures, Japanese children have had massive exposure to a social and ethical curriculum emphasizing responsibility, collaboration, kindness, and so forth (Easley & Easley, 1983; Lewis, 1995; Peak, 1991). Comparative survey research suggests that, by fifth grade, Japanese students have internalized school rules more strongly than have American students, and that American students are more likely than Japanese student to cite external reasons to learn — such as grades or adult expectations (Beaton et al., 1996; Hamilton, Blumenfeld, Akoh, & Miura, 1989a, 1989b; Schmidt et al., 1997).

In addition to the activities undertaken by Japanese schools, Japan se_families and larger society are also well positioned to promote children’s positive emotional attachment to school and commitment to values of hard work, cooperation, and so forth. For example, rates of poverty, divorce, single parenthood, and school mobility are all very low by U.S. standards (Ladd, 1995).

Organization of Schools for Student and Adult Learning

In Japanese elementary schools, the school day typically consists of five or six 45-minute class periods, each followed by a recess of 10-20 minutes during which children play very actively (Lewis, 1995; Stevenson & Stigler, 1992). Students remain together all day for 2 years as a class of up to 40 students. A total of 105 periods per year (amounting to 2-3 periods per week) are devoted to science instruction, which starts in Grade 3. Comparative studies of U.S. and Japanese classrooms reveal that Japanese mathematics lessons are rarely interrupted by announcements or by pullout of students, whereas such disruptions occur regularly in the U.S. (Stigler & Hiebert, 1997), a pattern that likely applies to science instruction as well. The uninterrupted 45-minute lesson blocks may both reflect beliefs about the importance of deep "sticky-probing" of students' thinking (Hess & Azuma, 1991) and lesson coherence (Stigler et al., 1996) as well as create the conditions for these to occur. In addition, the 45-minute blocks make Japanese lessons predictable to students, who know how much time is available and can pace themselves through activities such as experimentation and write-up (Tsuchida & Lewis, 1996).

Despite the fact that most Japanese elementary teachers are responsible for all subjects, they typically introduce themselves as specialists in a particular subject matter (such as science) which they pursue as the main focus of their own professional development and their leadership within the school.

One misunderstood feature of Japanese elementary education is the time teachers spend engaged with students, which is generally comparable to the U.S.: from 8:30 a.m. until mid-afternoon or later, depending on club responsibilities (see data from several school districts presented in Lewis, 1995). The misunderstanding seems to stem from accounts that cite only the 25 periods (19 hours) per week of required elementary instruction, ignoring the additional 90 minutes or so (7 to 8 hours per week) teachers spend each day in the classroom with students, eating lunch with them, cleaning the classroom, and participating in twice-daily class meetings (Lewis, 1995). Many teachers have the additional responsibility of after-school clubs and school-wide special activities.

Japanese elementary teachers collaborate routinely: They plan and reflect on lessons and regularly present research lessons (kenkyuu jugyou) to colleagues (Lewis & Tsuchida, 1997; Shimahara & Sakai, 1995; Stigler & Hiebert, 1997). In the course of these activities, they frequently discuss alternate approaches to content: for example, the advantages of tuning forks versus stringed instruments to introduce Sound and vibration. They also collaborate on "research lessons" that are watched by all their colleagues, and which bring to life new instructional emphases such as fostering Student "active problem-solving," or "desire to learn" (Lewis Tsuchida, 1997). In these research lessons, teachers jointly Plan (often for months) how to improve some aspect of curriculum or practice, invite the entire faculty to observe and record the lesson, and then discuss it (Lewis & Tsuchida, 1997). Excerpts from research lessons are included below in our discussion of the connections between science activity structures and system features.

The Course of Study Established Through National Guidelines, Textbooks, and Policy

Although commentators frequently speak of Japan's "national curriculum," what actually exists is one slim volume of general goals for elementary education, The Course of Study for Elementary Schools, and an additional slim guide for each subject area.

To be approved for use, textbooks must cover the content laid out in these volumes. For elementary science, six approved textbook series compete for adoption by local school districts. Because textbooks translate the national Course of Study into classroom lessons, teachers who follow the textbook know they are covering the national requirements. Yet, like the Course of Study, textbooks are spare. For example, looking at two of the approved textbook series, we find that the entire fourth-grade electricity unit, which normally takes 10 class periods, covers just 12 to 13 pages, with most of the space devoted to colorful illustrations of students conducting experiments, and just one tenth the text contained in U.S. electricity units. The textbooks do not just happen to be spare in content. Textbooks are expected to meet a very modest page limit (e.g., 120 six-by-ten-inch pages for the sixth-grade science text), and are disapproved if they go beyond that included in the Course of Study -- for example, if students are asked to link up three batteries instead of two (Nagano, 1997).

TIMSS analyses suggest that the Japanese curriculum is highly focused and coherent (i.e., includes fewer topics taught per year and more time devoted to each topic than do curricula in many other countries). For example, of the 105 forty-five-minute lessons per year, fifth graders spend about 12 lessons studying levers, and fourth graders spend 13 lessons on dissolving (as shown in Table 2). Japanese eighth-grade science textbooks cover eight topics, compared to an average of more than 65 for U.S. eighth-grade textbooks (Schmidt et al., 1997).

What are the Connections Between Science Activity Structures and System Features?

In addition to the consistency of the science activity structures listed in Table 1, we were also struck by the interdependence of these science activity structures with educational system features. For example, students shared fragile and potentially dangerous materials in small groups, conducting experiments without major accidents or disputes, and cleaned up spills such as broken glass and eggs with-out instruction from the teacher. They used matches and flames, oxygen canisters, and oxygen detectors whose glass tips had to be broken for activation as well as long, heavy steel poles to lift pails of sand. We wondered if these materials could have been used successfully by classes of 30-40 elementary schoolers under the supervision of one adult if the students had not been trained, from early in their school careers, to work collaboratively and responsibly in small groups (Lewis, 1995).

Another way in which science activity structures may depend on system features is the various habits and procedures that are developed during the preschool and elementary years (Lewis, 1995; Peak, 1991), such as taking notes, self-managing transitions, and remembering required possessions (books, notebooks, pencils, colored pencils, glue-stick, etc.). We saw students spontaneously begin to record information in their notebooks without any teacher instruction on numerous occasions. On other occasions teachers provided handouts that guided students’ activities. In these cases students often spontaneously incorporated their completed handouts into their notebooks by gluing them onto the pages (using glue-sticks they carried). Transitions between whole-class discussion and small-group activity occurred in all the lessons we observed (the most common pattern was to begin and end with whole-class discussion with small-group and/or individual activity in between); students managed these transitions skillfully and with no direct intervention from the teacher even when they required carrying heavy or fragile equipment around the room. Teachers did not use rewards, punishments, or threats; they appeared to rely on the commitment and skills developed by students in the course of earlier schooling.

Discussion (a feature of 9 of the 10 lessons we saw) also appeared to benefit from established system features. Small group discussion was typically quite lively and students expected others to justify their views. In one group that we observed, two girls and two boys discussed how to investigate whether any oxygen or carbon dioxide was left in a closed jar after the candle in the jar burned out. The two girls argued that the burning of the candle in the jar had produced carbon dioxide. The two boys, on the other hand, argued that the carbon dioxide was originally contained in the air before the candle went out. The girls tried to win the boys over to their view by reminding them how humans inhale oxygen and exhale carbon dioxide. In the midst of this heated discussion, the teacher said to the group: "You should not be arguing so much about what you believe at this time; instead, write down how you can investigate what you're arguing about."

During whole-class discussions, students often expressed their opinions in a Systematic way that linked them to other students’ comments. For example, students in several classrooms used a set of hand gestures (e.g., an open palm, a fist, or two fingers in a "v") to indicate their wish to agree or disagree with others, or to elaborate on others' opinions. Often students prefaced their remarks by expressing their agreement or disagreement with other students, or connecting their ideas to others' ideas, for example: "I object to the experiments mentioned by Group 1 a little bit. . . " "My idea is similar to Sano-san's idea but. . . " or "In general, I support Nakamura-kun's idea …"

Several conditions may provide an infrastructure for lively discussion in Japan. Japanese teachers, more often than their American counterparts, encourage students to express their agreement or disagreement with other students during science lessons (Tsuchida, 1993; Tsuchida & Lewis, 1996), and use a slow-paced, "sticky-probe" approach that uncovers inconsistencies and avoids early consensus (Hess & Azuma, 1991). Sharing mistakes appears to be common in Japenese instruction, so that Japanese students are likely to become comfortable expressing an idea that eventually turns out to be wrong (Lewis, 1995; Stevenson & Stigler, 1992). Another sort of support for lively discussion may come from the belief that one's ideas will be treated with respect by classmates Research conducted in U.S. classrooms indicates that the benefits of small-group learning depend upon the social quality of children's interaction in groups: When students report that interaction in groups is respectful and inclusive, greater use of small groups is associated with increases in academic and prosocial outcomes; when group interaction is disrespectful and inequitable, greater use of groups is actually associated with reduced academic and prosocial outcomes (Battistich, Solomon, & Delucchi, 19903; Cohen, 1984; Linn & Burbules 1993).

The social and ethical curriculum experienced during the early school years in Japan may support the science instruction techniques we observed. For example, the lively discussions we saw in the course of many lessons may depend upon teachers' long-term efforts to build family like small groups where students feel comfortable raising ideas and responding to one another (Lewis, 1995).

Vigorous discussion may also depend on the teacher's conception of science as being "about" students' ideas (a concept strongly supported by Japan's Course of Study), rather than only about students experimenting on ideas offered by the teacher. For example, a Japanese teacher reflected, when asked how his science teaching, had changed over the eight times he had taught the combustion unit in his 16-year teaching career:

"My teaching has changed over the years because my own concept of science has changed. In my early years of teaching, I believed that science simply meant having students experiment. But in recent years, I've realized that ... children form their own concepts, form their own ideas about science, including the invisible world of science.... I am more willing to let students pursue what they want to investigate. As a result, I think that the children themselves have changed and are more able to take initiative in deciding what they want to investigate and expressing it.

Likewise, the science activity of eliciting students' prior knowledge-an approach widely advocated in U.S. model science programs (Eylon & Linn, 1988) -- may derive support from Japanese expectations of universal student participation. As Tsuchida and Lewis (1996) have pointed out, volunteering one's -ideas is not so much an individual choice as an obligation in Japanese elementary classrooms, with many techniques supporting participation by every class member. For example, teachers often go up and down rows to elicit every student's ideas about a topic; all students' ideas may be represented on the blackboard by magnets with each student's name, which are moved around to show chances in students' predictions; if student work is posted, it is the work of every student in the class; students are often asked to publicly debate and "vote" on their predictions about an experiment. Reciprocally, questions such "Did I raise my hand sometime today?" and "Are we a class in which all students speak up actively, or do just a few students speak up?" are common prompts offered by teachers for student reflection (Lewis, 1995).

Professional development is yet anoterh system feature that may support the science activity structures we observed. While Japanese teachers ultimately choose the instructional approaches they will use in the classroom, shared research lessons may offer opportunities for teachers to collectively build and refine not just instructional techniques, but also norms about what is good instruction. Field notes from Lesson2, a fifth-grade lesson on dissolving substances in water illustrate this:

The teacher, Mr. Oka, had prepared a handout listing students' prior experiments and findings, with findings from similar experiments grouped near one another. Mr. Oka asked the students to identify similar findings, connect them with arrows, and write brief explanations for the similarities. To further clarify what he meant by "connect similar findings," Mr. Oka posted two sentence strips, one at a time, on the blackboard. The first sentence strip said, "Mr. Oka is an animal." The second strip said, "Mr. Oka looks Like a monkey." Immediately, the students started laughing loudly. Over the din, Mr. Oka asked students to use arrows to connect their findings to similar findings from other groups. After about 5 minutes, Mr. Oka posted a third sentence strip that said, "Humans are in the monkey family." Forming a triangle-like shape with the three sentence strips, Mr. Oka asked, "Can you think of an arrow to connect the first and third sentence strips, and think about the reason why they are connected to one another?" Most of the students had a hard time connecting the monkey analogy to the task that they were supposed to perform. The lack of connection between the teacher's analogy and students' task and the resulting confusion among, students was obvious to other teachers observing the lesson.

In the discussion following the lesson, the following conversation occurred:

Visiting Teacher: Why did you decide to use the monkey analogy? It didn't provide the students with the foundation needed to build associations; instead it confused them. You should have provided a more relevant example, using findings from the students' own experiments, or at least an example directly relevant to the topic of dissolving. Since students didn't understand the purpose of the task, they simply connected the words on the paper without really analyzing the relationships behind the connections.

Oka: First of all, I didn't think that the example was good either. However, since it was the first time for the students to connect findings, the goal was just to look for the point of connection and the reason. Thus, I had to hold myself back. That is why I came up with a poor example.

Observer: So in other words, what's important is not so much for students to connect the findings, but to think carefully about how they relate to each other. Is that right? That's where the focus of your lesson was?

Oka: Yes.

The example illustrates the opportunity for Japanese teachers to improve their practice by having their lessons observed and critiqued by other teachers (see also Lewis & Tsuchida, 1997). It points out that even experienced Japanese teachers like Mr. Oka, at national schools that to some extent serve as "lighthouses" for other elementary schools, sometimes choose strategies that fail to engage students' understanding.

The comments of a local public school principal point out other system features related to professional development, including the importance of teacher-led (rather than top down) instructional improvement, and the belief in shared, public lessons as a catalyst for improvement:

The ward made this school a research school for science. So people think that the teachers in this school are ... superior at science. To be honest with you, we don't have, good science teachers here. No one is good at science here. So, why and how will they do research? It is necessary for teachers themselves to think about how to teach science, to tell their ideas frankly to others, to ... improve themselves. The teachers in this school don't know much about science, but with their own knowledge, they will express their opinions as to what kind of lessons they want to do and what kind of teaching materials they want to develop within this school. Since this school is ... the site for National Science Teachers' Association conference, teachers from other schools in Tokyo will assist this school, because they want the research lessons at the Tokyo conference to be successful. Based upon their exchange Of opinions, our staff will redesign lesson plans or redevelop teaching materials. Then, they will conduct the research lessons. Our staff and those teachers who assist our staff ... will improve themselves together. That is how we work together.

Although the principal was himself a science specialist, he did not use his own expertise to improve the school's science instruction; rather, he emphasized peer collaboration, in which teachers not expert at teaching science had the chance to express their own ideas and plans to more expert colleagues, and to learn, over time, through discussion and trial of actual classroom lessons. In its emphasis on all teachers expressing their ideas and plans, sharing information with colleagues, and conducting trials (in the form of classroom lessons), the model of teachers learning is similar to that of student science learning (see Table 1). Teachers, thus, gained personal experience with the type of instruction they were trying to build in their classrooms. Another synergy between system features and science activity structures suggested by the principal's comments is that research lessons may create a "demand" to improve practice, because lessons will be viewed by other teachers (Lewis & Tsuchicla, 1997).

Teachers' peer collaboration also reinforces shared activity structures. For example, after a research lesson on dissolving (observed by Tsuchida on February 8, 1996), several teachers (who happened to be younger teachers) advocated letting students dissolve any materials familiar to them, such as sugar and powdered juice. These teachers argued that using familiar materials would strongly motivate students to investigate the nature of water solutions, and would facilitate students bringing their daily life experience into science lessons. In contrast, several more experienced teachers opposed the idea and thought that teachers should use only materials that would demonstrate a clear threshold of solubility in water - materials such as salt and boric acid, suggested by the textbook. Otherwise, these teachers argued, students would miss a key point of the lesson: that some materials remain solid when put in water. The first group, on the other hand, criticized the latter group for "forcing students to do the experiment that teachers planned, rather than letting them conduct an experiment from their own pure interest and curiosity."

The exchanges, like those just quoted, that take place at research lessons might be thought of as "capacity-building" conversations: Teachers are developing and refining their ideas of what good instruction is, their knowledge of supporting techniques, and their judgment about when to emphasize teacher-directed versus student-initiated approaches. The last quality - development of judgment - may be a support for science lessons that is more readily fostered by shared discussion of real classroom lessons than by the types of professional development more widely available in the U.S., such as one-way dissemination of new approaches, or adult learning that occurs without children present.

The effectiveness of research lessons may depend, in turn, on another system feature: A shared national curriculum that focuses in depth on a small number of topics. For example, a teachers manual outlines the study of dissolving as shown in Table 2. It illustrates that the curriculum allows time, in the course of 13 lessons on dissolving, to use the science activity structures shown in Table 1: To elicit students' ideas, conduct experiments, have students exchange information, revisit initial ideas, and so forth. Because the curriculum is shared, teachers can have a focused dialogue on the strategies, purposes, and fine points of teaching a unit.

Discussion

While definitive conclusions about the nature of Japanese elementary science instruction await large-scale, representative studies, data from the small, convenience samples studied to date suggest that the science activity structures found in Japanese elementary lessons may overlap substantially with those found in many model programs in the United States. Indeed, Japanese elementary teachers expressed surprise that we were so interested in their science instruction, which they saw as heavily influenced by Western approaches, including the work of John Dewey and Jerome Bruner, discovery learning, inquiry-based approaches, and various Sputnik-inspired reforms. Indeed, the activities for the study of dissolving (Table 2) could have come from the Science Curriculum Improvement Study (Karplus & Thier, 1967). Our observations raise the possibility that a set of science activity structures may be implemented systematically and repeatedly over man), lessons and man), years in Japan, thereby building coherent science instruction.

Our observations also suggest several ways that Japanese elementary science activity structures may depend upon features of the larger educational system. For example, a strong focus on collaboration and personal responsibility during the preschool and elementary years may build skills and attitudes essential to the vigorous, yet respectful conversations likely to promote learning (Battistich et al., 1993; Cohen, 1984). Collaborative, lesson-based professional development mav be essential for teachers to develop, hone, and spread science activity structures. Japan's frugal curriculum0lives teachers time to use student-centered techniques, such as eliciting ideas and revisiting hypotheses; the shared curriculum gives teachers natural grounds for collaboration.

We draw several implications from our integration of observational data and research on Japanese system features. First, research is needed that identifies the interdependencies between science activity structures and educational system features. Our review suggests that, by late elementary school, the Japanese classroom context may differ substantially from that of the U.S. in many aspects, including student attitudes and skills, opportunities for collaborative learning by both teachers and students, and pressure for curriculum coverage. It should not be assumed, therefore, that science activity structures used by Japanese teachers could be used successfully in U.S. settings.

For example, a science activity structure such as student information exchange or debate may derive support from the long-term social and ethical emphases of Japanese schooling, in which respectful, lively discussion in family like small groups is nurtured from preschool on (Lewis, 1995; Peak, 1991). It is questionable whether Japanese techniques such as highlighting and discussing students' erroneous predictions would be effective in settings where teachers had not worked to create a supportive, family-like classroom culture. As noted earlier, evidence from U.S. schools suggests that cooperative learning benefits students only when group conversations are respectful and inclusive. Similarly, the capacity of Japanese students to work independently, conducting experiments using potentially dangerous materials, may depend on substantial efforts to build personal responsibility and internalized motivation.

Second, our work suggests the need to study science education (and perhaps academic instruction of all kinds) in the larger context of students' social and ethical development. The science activity structures used in Japanese schools may depend, for their success, on dimensions of students' social and ethical development, such as responsibility, helpfulness, and the willingness and capacity to express disagreement respectfully. Yet few reform efforts in the U.S. focus simultaneously on academic content, social development, and character development (Lewis, Schaps, & Watson, 1995).

Finally, if the finding of consistency in science activity structures is borne out by future research, then it is important to document the system features that support a wide range of teachers in using, these science activity structures coherently. The research lessons we observed suggest that Japanese teachers have the opportunity to analyze and improve science activity structures as they see each other's lessons and exchange reactions. Although much research suggests that model instructional approaches tend to lose important features, and even become empty shells, when disseminated to new sites, relatively little research explores the conditions that contribute to improvement of model approaches as they are spread. The science specialist whose confusing analogy was pointed out by colleagues reminds us that Japanese practitioners do not spring fully equipped into the classroom; they make mistakes and, it seems, have opportunities to learn from them.

In the United States, researchers are beginning to recognize the essential value of teachers' beliefs and experiences in the implementation of reform efforts, leading to new models of professional development which similarly use teaching practice as creative new learning opportunities for teachers (Ball & Cohen, 1996; Putnam & Borko, 1997). In one research program, Kids as Global Scientists, teacher professional development is situated in the context of teaching a common curriculum which is differentially enacted in hundreds of classrooms simultaneously while it also becomes the object of shared electronic discourse. As implementation occurs, both experienced teachers and project newcomers work with and reflect upon peer-generated cases or issues that arise during varied classroom experiences (Songer, 1998). While many questions remain as to how to best capitalize on these enacted learning opportunities for reflection and understanding, it behooves us to begin an examination of the common insights gained from observations of Japanese professional development and our own work leading to an expanded understanding of these complex issues.

What kinds of instructional approaches are most amenable to being honed and improved as they are disseminated, and what conditions are needed for such improvement to occur? For example, does improvement depend upon a frugal curriculum. that provides time for active, in-depth learning, so that teachers are free to focus on how to teach, rather than what to teach? Does it depend on a shared curriculum? On a teacher-led professional development system in which teachers routinely watch each other's classroom lessons?

The data on achievement, curriculum, and classroom practice gathered as part of TIMSS provide an unprecedented resource for identifying some system features in the countries that most effectively educate their students in science and mathematics; we hope they will catalyze further study of the connections between classroom activity structures and features of the lar-er educational system. When we show videotapes of Japanese elementary science lessons to American educators, two reactions are common: to note the many similarities between Japanese practices and U.S. model science programs; and to say, "That kind of instruction could never happen here." Both reactions suggest productive questions for future research: What are the conditions that enable Japanese teachers to spread and hone effective instructional practices? And what are the conditions needed to support broad use of effective instructional practices?

(References must be obtained from the original manuscript)