Designing a Model-Based Methodology for Science Instruction:
Lessons from a Bilingual Classroom

Cory Buxton
University of New Orleans

Abstract

This study reports on findings from the “Science Theater/Teatro de Ciencias” (sTc) project. The goal of sTc was to explore the potential of using student-generated computer models as a medium for elementary school students to develop richer and more meaningful explanations of science content. A secondary goal was to effectively engage culturally and linguistically diverse students in science learning. This paper reports on findings from a second/third grade two-way bilingual classroom. Conceptually, I rely upon a sociocultural perspective that differs from previous work that has been done using computer models in science classrooms. Specifically, I explore two issues that expand upon this prior work: 1) the potential value of using computer modeling for science learning in the primary elementary grades; and 2) the role that clarifying one’s personal understanding of how science is practiced plays in students’ academic success in school science. My assertion is that model-based science instruction can provide an effective strategy for mediating the barriers to success in school science that CLD students (as well as many other students) often encounter.

Sections of the Article

The focus of this article is on how the students in one experimental classroom (a 2nd/3rd-grade two-way bilingual [English-Spanish] classroom in a small western city) learned to negotiate a greater understanding of scientific practice through interactions with their peers, their teachers, and the researchers. This was done through the collaborative development of a model-based methodology for science instruction in which students created computer models in conjunction with the construction of physical models, other hands-on activities, and ample time to talk about these constructs in the classroom setting. My basic thesis is that such a model-based methodology for science instruction can provide an effective strategy for mediating the barriers to success in school science that CLD students (as well as many other students) often encounter.

Computer Modeling and Science Learning

The role of instructional technology, and more specifically the role of computers as tools for enhancing science learning, has been a topic of a great deal of study since the rise of the microcomputer in the 1980s. As Marcia Linn argued over a decade ago (1987),

This quote is just as applicable today as it was in 1987. While we have learned a great deal about computers as tools for science teaching and learning, there is still much more to be discovered. More recently, Linn (1992) has explored three effective uses for computers in classroom settings: 1) to collect and display data that students can then interpret; 2) to communicate and synthesize information from a number of sources situated across space and time; and 3) to help students predict and test natural phenomena by asking members to make predictions, agree or disagree with the group consensus, and then test those predictions using computer models. Unfortunately, computers are seldom used in these ways in elementary school science.

Much of the innovative research that has been done on uses of computers in the science classroom falls into Linn's first category, using computers to collect and display data for students to interpret. This type of project has come to be known as Microcomputer-Based Laboratories (MBL's) (Linn, Layman, & Nachmias, 1987; Mokros & Tinker, 1987; Reif, 1987; Rivers & Vockell, 1987). Studies using MBL's have been done almost exclusively at the secondary school level.

The present study falls within Linn's third category of effective computer use, getting students to make predictions about natural phenomena and then testing or exploring those predictions using computer models. Much of the other salient work in this category has been done by White and Frederiksen (1999, 1990). They have explored model-based problem solving in both secondary and post-secondary educational settings, focusing on how students develop what they refer to as Intermediate Causal Models (ICM's) and how these models can then be used as problem solving tools. White and Frederiksen claim that there are four instructional ramifications for their work with student-generated computer modeling. They are to: 1) use semi-abstract representations in modeling physical systems; 2) teach modeling, not just models; 3) teach problem solving as model-based reasoning; and 4) recognize the critical importance of metacognition.

As can be seen from the examples presented in this section, most of the work done to date using computers as tools to aid students' science learning has been decidedly cognitive in nature. These studies have focused on identifying and ameliorating students' misconceptions, and developing their abilities in problem solving and critical thinking, as well as creating a better understanding of the processes used in the practice of science.

This prior work has been done almost exclusively in secondary school settings. In contrast, little has been done with computer models that focus on the social construction of scientific understanding or how the creation and use of student-generated computer models can serve as a tool for students to shape their understanding of the meaning of science. Thus, the sTc project expanded on the prior work done in this area in two ways: 1) by pushing the question of appropriate use of computers for science learning down to the primary elementary grades; and 2) by focusing on the role that clarifying one's personal understanding of how science is practiced plays in students' academic success in school science.

The Setting

The sTc project took place at Front Range Elementary, a mid-sized elementary school in a small western city. The school district had an open enrollment policy, meaning that any student could attend any school, a policy that gave rise to a number of focus programs. Front Range Elementary was one of two elementary schools in the district offering a two-way bilingual program. In this setting, classes were composed of roughly half-native English speakers and half-native Spanish speakers. Instruction alternated by day, so that on one day all instruction (including science) was in English, and the next day, all instruction (again including science) was in Spanish.

Overall, the sTc project took place over a three-year period (1995-1998) in two classrooms at Front Range Elementary. Data for this article came from only one of those classrooms, a combined second and third grade class. Over the three years of the study, this class averaged 26 students, 60% of whom were native Spanish speakers and 40% of whom were native English speakers. The classroom teacher, Teresa Garcia, was a veteran teacher with 20 years experience working in bilingual programs and 14 years at Front Range Elementary. A bilingual classroom aide, Flora Gutierrez, was also present in the classroom for most of the school day.

Data Collection

The majority of data for this particular study was collected during the 1997-98 academic year. During the year the class worked on four science units, each about eight weeks in length. The school followed the practice of picking a school-wide science theme each year, which for the year in question was "environments and ecosystems." This theme served as a constraint upon the topics that we taught during the year as well as the basis for a school-wide "Science Museum" held each spring. The first unit we developed focused on seed dispersal mechanisms, which we also used as an introduction to modeling and to the computer software. The second unit was pollination, the third was food webs, and the fourth unit was recycling. The project researchers took ethnographic field notes during classroom sessions twice weekly. During these sessions we functioned as participant observers, engaged at various times in teaching the entire class, working with small groups, working with individual students in the classroom and the computer lab, or stepping back to observe as Teresa lead the class.

Additional data came from the collection of classroom artifacts such as copies of student-produced computer models, stories and drawings, as well as books, handouts, and other instructional materials used in the class. Finally, individual interviews were conducted with the students at two points during the year (at the end of the Science Museum and at the end of the recycling unit), and periodically with the classroom teacher, Teresa.

Data Analysis

The Constant Comparative Method, originally developed by Glaser and Strauss (1967), was used to derive a grounded theory. This approach to data analysis was particularly appropriate for this study for two reasons. First, since most of the prior studies of computer models in science education have been informed by perspectives from cognitive psychology, the sociological perspective that informs grounded theory provided a fresh lens with which to look at the issue of student model building. Second, since this study's emphasis was on collaborative interaction and meaning-making, this focus fit well with the sociological paradigm that underlies the grounded theory perspective.

Findings

Analysis of the data yielded patterns of behavior that illuminate a grounded theory I refer to as a model-based methodology for science instruction. Referring to a method of instruction as a grounded theory may strike the reader as a bit odd. However, in the context of this project, with almost no prior work to guide us, we had not developed this methodology in advance, struggling instead throughout the project with the question of how to make science-based computer modeling meaningful for young children. This method of instruction that evolved over time developed out of a melange of circumstances and contextual features, including: 1) the experiences of the classroom teacher working with bilingual/bicultural grade school children; 2) the experiences of the researchers teaching science and computer skills to young children; 3) the demands of teaching modeling (both computer-based and physical) to children of this age group; and 4) perhaps most importantly, from watching this group of students interact, and taking their lead in terms of how they designed successful learning experiences for themselves.

Thus, the framework of an emerging grounded theory seemed to me to be a logical way to describe the evolution of our project. I do not claim that the method of instruction we constructed is the only way to effectively teach science through model building to elementary school students. However, the lessons we learned in this setting have lead us to develop a set of understandings that I suspect would be beneficial in any setting where models are used to teach science to young children. Furthermore, I believe this approach to be especially effective in working with CLD students in the elementary grades.

The model-based methodology we developed can be expressed in terms of a sociological paradigm with five components: 1) causal conditions that lead to the methodology; 2) properties of the methodology; 3) the contexts in which the methodology developed; 4) actions taken in response to this methodology; and 5) consequences resulting from those actions. An exploration of the dynamic interplay among these five constructs will lead me, in the following sections, to elucidate the primary characteristics of the model-based methodology for science instruction that we developed in this setting. In the last section, I will discuss the more general implications of this methodology and tie this work back to the prior work done on computer modeling for science understanding.

Causal Conditions for Successful Use of the Methodology

One of the keys to being able to use models successfully as tools for science learning is for students to gain an understanding of the value of models, and what a given model is able to (and not able to) show. This understanding is intimately linked to understanding how to use models in meaningful ways. By using models in meaningful ways, I mean focusing on those aspects of the model that are relevant for science learning. This is a skill that has to be taught explicitly, as children are likely to focus on aspects of the model that are extraneous to the science concepts.

While we began the project talking about a number of possible valid uses for models, students gravitated toward the image of models as props or tools that are used to tell a story. While this was somewhat different from our initial goal of using models to explore causal mechanisms, "model as storytelling device," similar to an animated storybook, became a metaphor the students readily adopted. This metaphor, in turn, helped increase students' willingness to engage in content-related discussions, both with their peers and with adults. The following fieldnote excerpt shows this reasoning:

Other causal conditions for successful use of the method of instruction clustered around the idea of increasing students' comfort level, both in their interactions with computers and in their discussions of scientific ideas. One aspect of increasing students' comfort level with computer modeling had to do with choosing computer software with which students could be successful. Computer modeling is a complex task, and in order to get the students to focus on what they were modeling and not just how they got the model to work, the right software was needed. In our third and final year of the project, we switched the software we used from Cocoa (Cypher & Smith, 1998), a simulation design program, to the Amazing Animation program (Cornish, 1994), with the aim of creating animations rather than simulations. Animations emphasize spatial and temporal relationships rather than cause and effect mechanisms. With this change in software, we saw a dramatic difference in how students interacted with and made use of their models. In the following excerpt, Teresa, the classroom teacher, discusses her feelings about switching software.

FIGURE 1: Belle's Food Web Model

Another way of increasing students' comfort with science was to begin with experiences that were familiar to them and then gradually expand beyond those experiences. Two concrete manifestations of this strategy were to begin our studies with a focus on the local environment, and to give students the freedom to use either their first or second language in describing and talking about their models. The following excerpt provides an example of the first of these strategies:

In summary, the causal conditions that we found to be necessary for students to begin to engage in meaningful model use were an understanding of the purpose of models and the development of a classroom environment where students became comfortable exploring what it meant to think, do, and talk about science. Factors that lead to an adequate student comfort level included selecting the appropriate computer software and starting with topics that students could relate to based on their own prior experiences.

Properties of the Methodology

The model-based method of instruction that we developed for teaching science in this setting evolved out of the causal conditions described above. While we were not able to enumerate these causal conditions from the outset of the project, by paying close attention to when students were able to succeed in the modeling tasks and when they struggled, we gradually became aware of the causal conditions and modified our tasks and strategies accordingly. So, for example, one of the properties of instruction that developed over time was an increased reliance on students' literacy and drawing skills in the classroom.

The project was originally designed to circumvent the need for student literacy due to the pictorial nature of the computer modeling programs that we selected. However, we found that activities done in the classroom prior to students starting to work on the computers (what we came to refer to as pre-modeling activities) were essential for guiding students' work when they began to make their models. These pre-modeling activities were predominantly literacy-based. However, since the range of literacy skills in the class was quite broad, we found that some students were more successful drawing pictures of what they planned to show in their model rather than writing out a narrative. The following excerpt demonstrates this approach:

Another property of the method of instruction was an emphasis on the scientific processes of design, experimentation, and revision. These three characteristics can be related to the three stages of modeling that we addressed: pre-modeling, modeling, and post-modeling. All three of these steps were important for the successful construction of models. We developed several activities in the classroom that made use of all three steps using physical models rather than computer models in order to teach these processes. The following is an example of such a physical modeling activity:

Another key property of the evolving method of instruction was related to providing the optimum amount of student choice in the process. As researchers we began the project with the belief that allowing a great degree of student choice and direction would lead to student ownership of the models and a better final product. Over time we found, however, that with young students and a complex project, frustration and a sense of being overwhelmed could quickly set in with negative consequences. We gradually began to provide more structure in the form of pre-modeling activities. These activities were meant to decrease the confusion that is described in the following excerpt:

In summary, the properties of our method of model-based science instruction can be thought of in terms of finding the proper balance along several dimensions. First is the balance among students' written expression, their oral expression, and their expression via computer models. While we started the project with the hope that the computer models themselves would be the medium of communication through which students would be able to express their conceptual understanding of science, we found that other modes of communication needed to be accessed before students could make successful models. Thus, we needed to balance the amount of pre-modeling that we required with the opportunities students had to create their models. Put another way, we tried to balance model design, model creation, and model revision. The other major factor to balance was the amount of student choice and adult direction. While we began the project giving students the maximum amount of choice, over time we began to provide more direction and guidance.

Context in Which the Methodology Occurred

In practice, our model-based methodology for science instruction differed in significant ways across the two principle settings of the project, the classroom and the computer lab. However, in both settings we gradually came to realize that we, as teachers and researchers, were not wholly responsible for the development of successful strategies. Discourse among students, both in the classroom and when at the computer, played a significant role in how they made sense both of their models, and of the science being represented therein. One common way that this understanding developed was through students talking about their own and each other's models as they made them. An example can be seen in the following passage:

FIGURE 2: Delilia's Prairie Dog Town

The above conversation shows how students' concerns often did not focus on the science content, but rather on the appearance of their models. The emphasis was on appearance, spatial relations, and on how to manipulate the program to do what the students wished. When students talked about their models on their own, they often focused on aspects that we did not view as the key considerations. Thus, in the above example the focus was on the size, color, and detail of the backgrounds, rather than on how the background would be used to demonstrate the concept of food webs. This was fairly typical of the discourse that took place when students were at the computers creating their models. However, this discourse changed markedly once the models had been finished. Talk about completed models generally focused on what the models were showing rather than on how they were constructed.

Within the classroom context, Teresa played a significant role in developing the methodology of instruction through her whole-class debriefing sessions after students had been working in small groups on pre-modeling activities.

Whole class discussions lead by one of the researchers were another instructional context used to introduce new kinds of activities. While periodic instruction in whole group sessions was valuable to get students thinking about a new activity or to refocus them on why they were doing what they had been doing, we also found that our normal conceptions of instruction had to change when so much of the students' work was done using computers. What we had to do to prepare students to make successful models differed in some ways from what we would have done to prepare them for other types of projects. Ample one-on-one interaction between students and adults was necessary for students to fully benefit from the project. These one-on-one interactions occurred both in the classroom setting and in the computer lab, and immerged as an important factor in our methodology of instruction. However, classroom discussions were generally about planning aspects of the models that were related to science content, whereas in the computer lab the clarifying discussions were most often related to how to do a specific task on the computer, and therefore, focused on computer content rather than science content. These two types of interactions can be seen in the following excerpts.

The kind of one-on-one tutoring described in these passages was significant in the students' developing ability to demonstrate their understanding of both the science concepts and the modeling process. Because our project allowed for the presence of several more adults working with students than is possible in the usual classroom setting, we were able to take particular advantage of this context.

In summary, the contexts in which the methodology of instruction was created fall into two basic categories: those that took place in the classroom and those that took place in the computer lab. Within the classroom setting there were distinctions among whole class discussions, small group activities, and one-on-one instruction. There were also distinctions between topics and activities focusing on science content and those focusing on model building. Within the computer lab setting the two most common contexts were peer interactions among students as they constructed their models and one-on-one tutoring with a researcher helping a student fix a particular problem with his or her model. These contexts can be directly linked to certain actions that took place in response to the methodology of instruction.

Actions Taken in Response to the Methodology of Instruction

Students engaged in a number of actions that indicated their growth and development both in terms of science content learning and the use of modeling as a way to express that learning. For example, as students became more proficient at designing their own models, they also learned to talk about their models to others. Students had to discuss their models at the end of each unit when we gave students a chance to present their model to a group of their peers in a show-and-tell format. There were two goals for these show-and-tells. First, they provided a chance for the model creator to express his or her understanding of both science concepts and model building by describing what his or her model showed and how it had been created. Second, the show-and-tells were meant to give students the opportunity to practice critiquing their peers' models and discuss ways in which a model could be revised to better express the science story that the designer was trying to tell. This critique and revision process is fundamental to how scientists use models in practice.

We found, however, that while these show-and-tells did fulfill the goal of getting students to describe their models in a safe context, students were much more hesitant to critique each others' models than they were to criticize their own. Perhaps because of Teresa's emphasis on building community and teaching students to respect one another, students seemed to feel compelled to compliment the designer and tell what they liked about the model, to the exclusion of talking about how the model could be improved. So, for example, the following frame from Josue's model of seed transport shows a common problem in many of the students' early models—objects floating in the air. However, as can be seen in the excerpt, during the show-and-tell portion of the activity no students in Josue's group mentioned the floating bear as something that could be improved.

FIGURE 3: Josue's Floating Bear

Throughout the project, one of our goals was to try to get students to add more detail to their science stories. We found that by the time students had completed all of the pre-modeling and model building activities in a given unit, they were generally able to explain the relevant science concepts in some detail. However, many students did not give these detailed descriptions spontaneously. Instead, they frequently needed adult prompting before they would give full explanations of what they knew. The following example demonstrates this in the case of the Science Museum display.

Another action that became common as students' familiarity with the modeling process increased was one student showing another student how to do a task on the computer. This was often done without adult prompting, but rather in response to a student's exclamations of frustration. Looking at what a neighbor was doing also served as a way in which students negotiated the meaning and the goals of their tasks. At other times, researchers had to remind students of how to do certain tasks.

In conclusion, several types of interesting actions arose from our methodology of science instruction. Perhaps most important were the pre-modeling activities that we found to be essential if young students were to be expected to create meaningful models at the computer. While these pre-modeling activities generally lead to reasonably successful models, we were never as successful in getting students to take actions to further improve their models through revision. Nor did students' models tend to reflect the same degree of complexity that they expressed in telling their science stories orally. We also found that actions differed significantly between the classroom setting and the computer lab setting, based on factors such as the complexity of the task, the amount of direction, and the opportunity to work collaboratively. The fifth and final component of our model is a consideration of the consequences of these actions.

Consequences of the Actions Taken

Many of the actions that students took in the process of creating their models resulted from the methodology of instruction that we developed in this classroom over time. In turn, these actions had consequences for how students came to conceive of both science and modeling making. One example is that, over time, students came to tell science stories based on their own experiences. While at the start of the project, students did not tend to relate science to their own lives, by the end, many seemed to do this naturally, as the following excerpt indicates:

This example clearly shows how students were learning to think about science in terms of their own experiences in the world around them and then link those experiences back to the science concepts that they were learning in class. Getting students to connect with science in personal ways was one of the primary goals of our project, and we all found examples of this to be extremely encouraging. However, from the classroom teacher's perspective, while Teresa was quite pleased with the science learning that was taking place, she also saw a certain cost to the implementation of our instructional approach. This cost had to do with the decreased flexibility that she had in letting the students' interests lead her choice of science content.

By participating in our project, Teresa basically agreed to trade off greater depth of topic coverage in exchange for less spontaneity in her teaching. Thus, one consequence of our methodology of instruction was that it placed certain constraints on how science could be taught. These constraints were directly related to the increased background that we found students needed on a given subject before they were able to construct a successful model that reflected some understanding of that topic. However, even when students had a grasp on the science content in question, there was still the additional layer of mastering the modeling software in order to create a successful representation.

Our time in the computer lab was generally spent working with students on the nuts and bolts of how to make the program do what they wanted it to, rather than on having students work with models in order to expand their understanding of the practice of science.

To facilitate students' interaction with science content through the medium of their models, we eventually shifted from an emphasis on causal mechanisms underlying science processes to an emphasis on how processes occur in time and space. Still, we had to constantly remind students that their computer modeling activities (both the actual work with the computers and the pre-modeling activities that took place in the classroom) were connected to the other science activities they were doing in the classroom with Teresa. One of the consequences of our actions, then, was that over time we gradually got students to consider the stories they were telling with their models in light of the relevant science content. The struggle to achieve this goal is clearly demonstrated in the following excerpt.

FIGURE 4: Oscar's Anthropomorphized Juggling Beavers

The interaction between Page and Oscar shows how conversations played a central role in getting students to connect the models they were making with the science content that they were studying. Because we knew that these conversations were important, we worked a great deal on getting the students to talk about and explain to others the science concepts they were learning. One thing we found was that the audience the students were talking to played a significant role in determining the detail in which the student was willing to discuss his or her model. In some cases students presented an accurate picture of what they knew about the topic, but in other cases they did not demonstrate what we knew from previous interactions that they understood.

Thus, finding the right audience to allow students to express what they had learned was critical for getting students to engage in meaningful scientific discussions. While we believed that Science Museum was an ideal setting for such discussions, we found that even then, other factors, such as the specific audience, influenced the effectiveness of the setting. Another factor related to audience had to do with the way in which students code switched when explaining or writing about the science concepts in the context of their model creation. The following two passages demonstrate some of the implications of code-switching in this context.

Two tendencies of second language acquisition can be seen in these passages. The first is that acquisition is rarely even between comprehension and production. Thus, while Kim was only able to fully comprehend the questions being asked in her L1 (Spanish), she felt comfortable enough responding to the questions in her L2 (English). Interestingly, this example runs counter to the generally accepted, but oversimplified notion that comprehension precedes production in the second language. The second point is that when learning technical vocabulary and concepts in the second language (such as Delilah and Julia discussing the recycling process with me in English), students often find that they do not then have the necessary vocabulary in their first language to write or talk about these concepts successfully. This may force them back into their second language.

In summary, the consequences of the methodology of science instruction we developed can be seen in terms of processes and situations that both helped and hindered students' attempts to express their evolving understanding of the practice of science. Factors that aided in students' ability to engage in scientific discourse included: 1) classroom discussions and activities aimed at helping students connect science to their own personal experiences; 2) greater depth of coverage of science topics; 3) the creation of settings such as the Science Museum and the show-and-tell sessions that encouraged students to talk about the science content demonstrated in their models; and 4) settings that forced students to deal with issues of first language-second language code switching in the discourse of science.

Factors that hindered students' ability to engage in scientific discourse included: 1) frustration that arose from the difficulty of mastering the nuances of the computer software; 2) the frequent need to focus on the "how to's" of model construction rather than on the science content being modeled when at the computers; and 3) the influence of some audiences that were not receptive to leaning science content from the students. In the final section, I now turn to more general implications of this study for teaching science to primary grade students in diverse settings.

Discussion and Implications

I have explored the development of our computer-based methodology of science instruction in terms of a five-faceted sociological paradigm of causal conditions, properties of the phenomenon, the contexts in which it occurred, actions used in response to the phenomenon, and consequences of those actions. From this analysis of our project, I believe that I can point to several lessons learned regarding the use of models to promote elementary students' success in learning to think about, act on, and talk about science. While these lessons obviously need to be tailored to the specific needs of any given setting, I would argue that they are sound principles for framing science instruction in the elementary classroom and are especially beneficial for culturally and linguistically diverse students. While by and large, these lessons are not new, they should be considered in light of the two novel goals of the sTc project, to push the use of computer-based modeling down to the middle elementary grades, and to focus on the role that gaining a personal understanding of the meaning of science plays in students' academic success in school science.

The first lesson is that successful elementary science students must come to understand the value and relevance of their own personal life experiences as a legitimate part of their understanding of science concepts. That is to say, they must come to believe that science is not just the purview of our socially constructed stereotypical scientist, the white man with disheveled hair and lab coat, mixing chemicals and deriving equations. In our study, we found that Teresa's role facilitating the whole class discussions provided students with ample opportunities to see the connections between their own experiences in the natural world and their understanding of science. Giving students autonomy over their model design also helped to reinforce students' beliefs in the value of their personal experiences.

A second lesson learned is that successful students must come to view the language of science as a discourse in which they can personally engage in an active manner. In other words, students must overcome their fear that the language of science is difficult, alien, and inaccessible to them. Using students' own models as a focus for the discourse helped to promote student comfort in talking about science both with peers and adults. However, we also found ample evidence that the use of computer models deflected some student talk away from science concepts and onto topics related to use of the computers themselves.

The third lesson learned about framing science instruction in the elementary classroom was the need to develop students' willingness to engage in content-related interactions with both peers and adults. This relates back to the issue of getting students to see themselves as evolving doers of science, rather than as empty receptacles waiting to be filled with science knowledge. In our study, many of the pre-modeling activities that students undertook to aid in conceptualizing the science stories they wished to tell with their models required students to ask questions, give advice, and collaboratively construct meaning with peers, researchers and the classroom teacher. The chocolate box model stands out as an especially good example of this.

A fourth lesson involved successful students learning to consider the social applications of science. When school science is removed from its social context, it becomes the isolated body of facts and "truths" that many of us were exposed to as students. This, in turn, causes many students, and especially culturally and linguistically diverse students, to feel disconnected and alienated from science. The topic selection in our project was a conscious attempt to emphasize the value of science as a means of addressing social problems. Issues such as local ecosystems and recycling were generally successful in sparking the interest and engagement of students.

The fifth and final lesson I will discuss here is that successful students must play an active role in helping the teacher tailor activities to their unique needs and abilities. Unlike the other computer-based research projects discussed in the beginning of this paper, our computer-based methodology was not something constructed by the researchers and then imposed on the students (and teacher). Rather, it was created collaboratively over time as students' desires and needs helped shape the instructional content and approach. Thus, while none of these lessons learned are likely to strike the reader as particularly novel, the successes and failures we had in bringing these ideas together through the use of student-generated computer models and implementing them in a culturally and linguistically diverse 2^nd and 3^rd grade classroom has left us with much food for thought.

As mentioned earlier, the sTc study expanded the boundaries of the previous research on using computers as tools for aiding in students' conceptual understanding of science both by pushing the question of appropriate use of computers for science learning down to the middle elementary grades, and by focusing on the need for students to come to terms with the meaning of science in personally meaningful ways.

In terms of the first issue, I believe that we have shown that using computer models with 2nd and 3rd graders can be beneficial to their developing abilities to think, act, and talk in ways that are compatible with the culture of school science. However, this may only be true in settings where a great deal of support, both technical and pedagogical, is provided. Thus, in our study, the presence of multiple researchers with substantial experience in both curriculum development and computer programming, allowed for the design of both pre-modeling and modeling activities which may not have been possible for Teresa alone to develop. Additionally, simply by adding several more adults to the room, we gave students access to a level of coaching and individualized attention that would not otherwise have been possible.

Even with all of this support we still were not always satisfied with what many of the students were able to produce. For example, the models students created generally did not demonstrate a degree of complexity equal to what the students actually knew about the topic (as assessed by formal interviews and informal conversations). Also, we engaged students in so many activities and discussions that they often had a hard time keeping all the ideas in their heads. As the units progressed, they seemed to learn new things while forgetting some of the prior concepts that they had known. Thus, models often disproportionately reflected the last topic that was addressed in pre-modeling activities, even when that topic was not the central issue. In short, even in an idealized learning environment, these 2nd and 3rd graders were pushed to their limits when asked to develop computer models and use those models as vehicles for thinking, acting, and talking in ways that reflect the practice of science.

The second implication, that students who gain a more personal understanding of science are more likely to be successful in school science, can be considered in light of the earlier claims made by White and Frederiksen. In our work on model building with significantly younger students, we found some consistencies but also significant differences with the work of White and Frederiksen. First, it should be noted that some differences in computer modeling ability are easily explained in terms of cognitive and perceptual developmental differences. For example, after two years of working with Intermediate Causal Models with our students, we concluded that the vast majority of these young students were not cognitively ready to design and create simulations of this kind, and we switched instead to animation-based models that were incapable of modeling causal mechanisms. However, we did conclude that both inquiry skills and the discourse of science could be adequately taught using animations rather than simulations. That said, in considering White and Frederiksen's four instructional ramifications, our data seem to support two of their claims and provide contrary evidence for the other two.

Our findings do not support White and Frederiksen's claim for the value of using semi-abstract representations in student models. Instead, we found that our students were better able to make use of concrete representations from their own personal experience as tools for problem solving. Our findings do support White and Frederiksen's second claim, namely, the importance of teaching modeling, not just models. Having students create their own models, through the process of design, creation, and revision activities was valuable to students both in their development of problem-solving skills and in the opportunities it provided for students to think, act, and talk like scientists.

We also saw evidence to support White and Frederiksen's third claim regarding the value of teaching problem solving in terms of model-based reasoning. However, this kind of reasoning was obviously more limited in our project since our use of animations did not provide the same opportunities to reason from the behaviors of a model to a scientific principle, as would be the case for simulation models. Finally, while cognitive and metacognitive skills play an obvious role in leaning to design and create computer models, I believe that in our study successful model creation can be attributed as much to social as to psychological processes. The development of students' science stories through group activities, and the refinement of these stories through peer interactions and discussions both during pre-modeling activities and while actually creating the models at the computers all point to the fundamental role of social interaction in the model building process. In other words, I see model building in the elementary science context as more of a social process than an individual cognitive task. This social construction of knowledge can then lead to a more personally connected understanding of the practice of science.

In conclusion, I believe that our study has shown that even for primary grade elementary students with limited prior exposure to computers, the use of student-generated computer models, in conjunction with the construction of physical models and other hands-on activities can provide meaningful opportunities for students to learn to think, act, and talk in accordance with the rules for success in school science settings. Experiences of this kind seemed to be especially valuable for culturally and linguistically diverse students who have historically been left out from meaningful science learning experiences. Further research on aspects such as how student discourse varies between computer-based and non-computer-based classroom settings and how students' knowledge transfers between these settings should provide greater insight into how computer modeling can be used most effectively in the science education of culturally and linguistically diverse learners.

Acknowledgements

I wish to thank the entire sTc project team: Clayton Lewis (Principal Investigator), Teresa Garcia, Carlos Garcia, Page Pulver, Cathy Brand, Cyndi Rader, Gina Cherry, & Linda Higgins. I would also like to thank the National Science Foundation, without whose support this project would not have been possible.

References

Buxton, C. (1998). Improving the science education of English Language Learners: Capitalizing on educational reform. Journal of Women and Minorities in Science and Engineering, 4 (4).

Chamot, A., & O'Malley, J. (1986). A cognitive academic language learning approach: An ESL content-based curriculum. Washington, DC: National Clearinghouse for Bilingual Education.

Cornish, A. (1994). Amazing Animation (Version 1.02). Modesto, CA: Vividus Corporation.

Cypher, A., & Smith, D. (1998). Cocoa (Version DR3). Cuperino, CA: Apple Computers.

Fathman, A., Quinn, M. E., & Kessler, C. (1992). Teaching science to English learners, Grades 4-8 (vol. 11). Washington, DC: National Clearinghouse for Bilingual Education.

Frederiksen, J., White, B., & Gutwill, J. (1999). Dynamic mental models in learning science: The importance of constructing derivational linkages among models. Journal of Research in Science Teaching, 36 (7), 806-836.

Glaser, B., & Strauss, A. (1967). The discovery of grounded theory: Strategies for qualitative research. Chicago, IL: Aldine Press.

Harding, S. (1991). Whose science? Whose knowledge? Thinking from women's lives. Ithaca: Cornell University Press.

Hawkins, J., & Pea, R. (1987). Tolls for bridging the cultures of everyday and scientific thinking. Journal of Research in Science Teaching, 24 (4), 291-307.

Kleinman, S. S. (1998). Overview of feminist perspectives on the ideology of science. Journal of Research in Science Teaching, 35 (8), 837-844.

Kohlstedt, S. G., & Longino, H. (1997). The women, gender and science question: What do research on women in science and research on gender and science have to do with each other? In S. G. Kohlstedt, & H. Longino (Eds.), Women, gender and science: New directions (vol. 12, pp. 3-15). Chicago: University of Chicago Press.

Krajcik, J., & Layman, J. (1993). Microcomputer-based laboratories in the science classroom. National Association for Research in Science Teaching, 35 (1), 3-6.

Lee, O., & Fradd, S. (1998). Science for all, including students from Non-English-Language-Backgrounds. Educational Researcher, 27 (4), 12-21.

Lemke, J. (1990). Talking science: Language, learning and values. Norwood, NJ: Ablex Publishing.

Linn, M. (1987). Guest editor's comments. Journal of Research in Science Teaching, 24 (4), 289-290.

Linn, M. (1992). Science education reform: Building on the research base. Journal of Research in Science Teaching, 29 (8), 821-840.

Linn, M., Layman, J., & Nachmias, R. (1987). Cognitive consequences of microcomputer-based laboratories: Graphing skills development. Contemporary Educational Psychology, 12, 244-253.

Longino, H. (1989). Can there be a feminist science? In N. Tuana (Ed.), Feminism and science (pp. 45-57). Bloomington, IN: Indiana University Press.

Longino, H. (1990). Science as social knowledge: Values and objectivity in scientific inquiry. Princeton, NJ: Princeton University Press.

Mason, C. L., & Barba, R. H. (1992). Equal opportunity science. Science Teacher, 59 (5), 23-26.

Mokros, J., & Tinker, R. (1987). The impact of microcomputer-based labs on children's ability to interpret graphs. Journal of Research in Science Teaching, 24 (4), 369-383.

National Research Council. (1996). National science education standards. Washington, DC: National Academy Press.

Reif, F. (1987). Instructional design, cognition, and technology: Applications to the teaching of scientific concepts. Journal of Research in Science Teaching, 24 (4), 309-324.

Rivers, R., & Vockell, E. (1987). Computer simulations to stimulate scientific problem solving. Journal of Research in Science Teaching, 24 (5), 403-415.

Rosenthal, J. W. (1996). Teaching science to language minority students: Theory and practice. London: Multilingual Matters, Ltd.

Salomon, G., Perkins, D., & Globerson, T. (1991). Partners in cognition: Extending human intelligence with intelligent technologies. Educational Researcher, 20 (3), 2-9.

White, B., & Frederiksen, J. (1990). Causal model progressions as a foundation for intelligent learning environments. Artificial Intelligence, 42, 99-157.