Introduction
The growth of web-based courses has generated a number of issues for lab science instructors. How do students satisfactorily complete the laboratory component of the course without coming to campus? Are home labs the equivalent of on-campus labs? Can simulations replace some or all of traditional hands-on lab? While these questions remain to be answered, the use of interactive, online simulations does appear to be one potential avenue for meeting lab instruction needs and improving student understanding and retention of physics concepts.

The concept of "learning by doing" is at the heart of both the behaviorists' and constructivists' support of lab instruction. Leonard noted that quality lab instruction contained at least three characteristics: student engagement in scientific inquiry, ability of students to manipulate the components, and a student experience designed to build understanding of the theoretical model (1989). The weakness of the traditional lab design is often in the prescribed nature of the lab exercise, which rarely includes any one of Leonard's criteria, especially building any connection between theory and application. The potential of simulations for assisting students to build mental models, transfer concepts, and apply theoretical knowledge to the physical world has been shown in a variety of studies (Lunetta & Hofstein, 1981; Choi & Genaro, 1987; Thomas & Hooper, 1991).

The recognition that student preconceptions form a barrier to conceptual understanding of physics principles as well as problem solving methods has revealed the need to re-evaluate the basic paradigms of physics education. Over the last 20 years, this has resulted in the development of a number of new tools and methods for instruction. Cognitive research on student learning in traditional physics education (Larkin, 1981; Clement, 1982; McDermott, 1984; Halloun & Hestenes, 1985; Mestre & Touger, 1989, Hestenes, et al., 1992) has generated a taxonomy of the preconceived misconceptions about the physical world, how these misconceptions interfere with student learning and strategies for overcoming these pre- and misconceptions within the topic of mechanics, especially kinematics. Physical models and definitions of motion are clearly not congruent with the preconceptions of motion that students generally have when entering a physics course. Preconceptions that interfere with student learning include explicit definitions of velocity (such as distance divided by time), no recognition of the difference between velocity and acceleration, no recognition that changes in direction imply acceleration, and that when an object reaches the top of its trajectory (momentarily coming to rest in the vertical direction) that no acceleration exists.

Several people have studied the differences between novice and expert problem solvers, especially the mental models each has of a given physical system, and persistence of the novices' naïve theories. One of the advantages held by experts is the "index" of experiences that assists the expert to readily identify the important information associated with a given problem system (Shank, Berman & Macpherson, 1999). How to get the novice student to recognize their misconceptions, tear down their mental constructs, and then build new mental models is the challenge.

Traditional physics instruction, supported by behaviorist theory, relies upon the teacher-centered method of lecture and example. This has proven to be ineffective at attending to the persistent preconceptions students have before entering a college level science course. On the other hand, constructivists' fundamental assumption is that an individual's knowledge is a consequence of their experiences and their construction of these experiences into body of knowledge. One of Mestre's conclusions is that it is important to provide students experiences that call into question their naïve theories. He advocates active participation of the learner in confronting and discussing contradictory explanation or predictions of physical phenomena (1991).

There is little dispute that the didactic methods by which physics has been taught are only effective for a specific class of learners, however efficient in terms of instructional time and cost they have been. Radically different approaches based on constructivist learning theory, although effective, are not practicable within the context and time constraints of current educational programs. Instruction that focuses on building causal models bridges the extremes. Teaching via causal models begins with simple representations of the system, and then directed activities guide students in applying causal reasoning to illustrate the mechanisms or models of the simple system at an intermediate level (White, 1993).
Instructional design for building causal models begins with the simple introduction of the concept via a "Source Model." The Source Model serves to introduce the concept through the most basic representations that are easily understood by or familiar to the students. Students require time to work with and master manipulating the Source Model to set up introduction of the "Derived Model." The Derived Model bridges the gap from simple to intermediate level conceptualization, and the Derived Model must be connected conceptually to the Source Model. The Derived Model must build upon the basic concept(s) of the Source Model, thus creating conceptual linkages between the two models and promoting transferability of the concept to generalized examples (Fredericksen, White & Gutwill, 1999).

Importance of Simulations
"Instructional computer simulations" are computer programs based upon mathematical models based upon either a physical model or theoretical prediction of the natural world that may be manipulated by the student user, often without the distractors present in most traditional representations (Weller, 1996). Thomas & Hooper noted that interactive programs allow manipulation from the introductory state through a number of intermediate steps to the final, or goal, state (1991).

Thomas & Hooper identified four applications for simulations in instruction:
· Experiencing simulations that prepare students for concepts to be examined in a traditional manner.
· Informing simulations used to augment or replace textbook instruction.
· Reinforcing simulations are designed to stay within the instructional context and reinforce learning.
· Integrating simulations bring together a set of separate facts or concepts and combine these into a global model of the systems studied.
Microworlds go beyond the minimally interactive simulations. These microworlds are based on an underlying set of mathematical models consistent with the global set of observable phenomena. The environment not only allows to interact with the simulation, but students are able to create their own simulations within the constraints of the global model (Weller, 1996).

Jonassen points out that interactive simulations are able to meet two of the three elements of student-centered learning environments (the other element being context). In addition to the simulation space, the interactivity of the simulations provides students the opportunity to manipulate the system beyond the traditional constraints (2000).

The use of microworlds, such as Interactive PhysicsTM, for assisting students in building conceptual models is supported by several theoretical perspectives (Jonassen, Peck & Wilson, 1999; Land & Hannafin, 2000). While simNewton does not have the range of capabilities of Interactive PhysicsTM, it does allow for the creation of web-based, manipulable simulations that can be used to develop and reinforce student's creation of causal models.

Simulations provide a potential solution for overcoming existing student misconceptions. Experiencing simulations that directly challenge misconceptions were found to influence students in recognizing the existence of problems in their personal "models" (Posner & Strike, 1989). Subsequent research by Flick (1990), Weller (1995), Henessey et al (1995 a, 1995b) and Gorsky and Finegold (1994) supported the use of interactive simulations as a tool for overcoming erroneous student preconceptions.

simNewton overview
simNewton is an interactive learning built on ThinkerTools, a simulation software package developed by White, Frederiksen, et al (1993). simNewton was created by Han Chin Liu (2002) to be a web-based implementation of ThinkerTools. In this environment, students see a "dot" representing an object moving, at first, without friction or gravity. They are asked to make the dot complete a task such as landing on an "x." The students can use the keyboard to apply a horizontal or vertical impulse, which is a hit or a force acting for a very short duration, to change the motion of the dot. The dot's trajectory can optionally be graphed as a dot print (rather than a foot print) path and the motion can be restrained with walls drawn to limit the motion of the dot. The simNewton lessons used follow White, et al's intermediate causal model. Each segment begins with simple representations of motion in one and two dimensions, and then builds to more complex representations tied to the initial simple representation of the concept. Activities progress from very simple tasks such as landing on the dot at a certain velocity, to landing a spaceship or guiding a massive object around a maze. Friction and gravity can be added in after the students have experience without them. Within the activities, the students make predictions of how the even will go, and then are asked to reflect on how the simulation matched their predictions. The student-controlled nature of the environment coupled with the student's ability to test various situations under constraints places it in the category of a microworld (Jonassen et al., 1999). The figure below shows some examples of the student interface for simNewton.

One of the simpler tasks	A more complicated task

The original ThinkerTools was produced in collaboration between the University of California, Berkeley, and the Educational Testing Service. It has developed into a comprehensive middle school curriculum for learning Newtonian mechanics and includes instructor materials, assessments and other materials for implementation in the classroom. The curriculum includes real activities intermingled with the simulations, so that the students can bridge their understanding between the real and virtual worlds. For example, before beginning the first simulation activity, the students are given a real croquet mallet and a ball. They practice hitting the ball so that they can deliver a unit of impulse to the ball consistently, measured by a sonic ranger type device or by measuring the distance the ball moves in the presence of friction such as on carpet. The goal is that the ball changes its velocity by the same amount for each hit. After achieving this, the students are asked to complete simple tasks similar to what they will do in the simulation, such as making the ball hit the target at a certain velocity. With this experience, then, the students interact with the simulations, and can transfer the real motion of the ball to the virtual motion of the dot. The predictive aspects of the scientific process are developed in depth in the curriculum. Students can make their own hypotheses and design and conduct virtual experiments to test them. Students draw their own conclusions base don't the evidence they see in the experiments.

Although ThinkerTools has been shown to be effective as described above (Whit, Fredericksen, 1999), Liu looked at whether a subset of the simulation activities of this type could be used independently and effectively. Liu created a package of these activities in simNewton, updating the interface to be web-based (2002). Liu's study tested simNewton with a group of middle school students and found these activities to be useful experiences in learning the basic principles of Newtonian motion.

In contrast to these previous studies, the present study observes the simNewton activities as they are used as part of the required curriculum for a college course in physical science, rather than with volunteer groups of middle school students.

Previous studies have shown that the order of activities in the unit affects student outcomes. For instance, Brant, Hooper and Sugrue (1991) found that using simulations before didactic instruction improved student scores. Studies conducted specifically using ThinkerTools gave similar results. This is supported by several learning theories that suggest that students must have an experience base from which to draw before being able to use symbolic models of the phenomena. (Andre et al., 1996) In the context of online classes, this offers a challenge to instructors as to how to guide the order of activities. Students are able to choose the order or their study activities to a great extent and in spite of encouragement from the instructor may complete the components of a unit in a different order. Course management software allows some control of the order that students can access materials, but this varies from platform to platform. The current study uses such controls to direct the order in which students complete the lab activity and text-based activities in an online course.

The Experiment
This investigation took place at a medium-sized community college located in a large mid-western, metropolitan area. Students were enrolled in an online physical science course with no pre-requisite requirements. The instructional goals of this course focus on building a conceptual understanding so students are not assessed on traditional problem solving ability.
Student access to the course is made through the WebCTTM course management system, and the designers have taken advantage of the conditional function in WebCTTM's testing and surveying tools to ensure that the treatment groups followed their instructional paths. Before beginning instruction, the students in the treatment groups completed a pre-test assessing their preconceptions. Submission of the pre-test was the trigger that released the instructional components of the course. The first treatment path has the students complete a review of the lecture notes, homework, and discussion activities before completing survey assessing their epistemological beliefs. Although this data is not analyzed here, completion of the survey was used to trigger the opening of the lab exercise link. The second treatment reversed the activities before and after the epistemological belief survey. The control format released all instructional elements simultaneously, and it used a traditional lab exercise using cars rolling down a slight incline to measure acceleration.

At the conclusion of the instructional module all students completed a post-test to assess student learning. The pre-test used questions from the Halloun & Hestenes assessment (1984), the student attitude survey is based upon a tool designed by Schommer (1993), and the post-test included content questions from the original course materials and the Force Concept Inventory (Hestenes et al., 1992).(Schommer, 1993)

Hypotheses-
o Students that undergo either of the treatment paths will perform better on the conceptual assessments than the control group.
o The two treatment paths will not demonstrate a significant difference in their performance on the assessment.

Data Analysis
The first two experimental groups (simNewton before content, SBC, and simNewton after content, SAC) completed the treatment during the spring semester while the control group, CON, completed the traditional lab with the content during the fall.

Pre-test
All students completed a pre-test consisting of ten questions directly addressing fundamental motion from the Force Concept Inventory prior to the release of content and lab exercises (possible using the conditional controls within the course management system). An initial test for Homogeneity using Levene's test, F(5,55)=..378, p > 0.05, and the kertosis, 1.246 and skew, -0.547, of the data illustrated that the data set fit the general assumption.

No significant difference was found between the three groups (CON, M = 5.00, SD = 1.612; SBC, M = 5.33, SD = 1.455; SAC, M = 5.59, SD = 2.282), F(2,50) = 0.021, p > 0.05. Analysis for temporal influences were not significant, F(1,60) < 0.001, p > 0.05, eliminating concerns associated with the delay between course offerings. Gender analysis was not possible as the subjects in all cases were skewed towards the female (CON, N = 21, male = 2; SBC, N = 18, male = 2; SAC, N = 22, male = 2).

A test for correlation was completed between the scores of the pre-test and that of the post-test (Module 2 Quiz) and very little correlation was found. When the same test was completed on the reduced set of questions (explained below) the result was similar (R2 = 0.0316). The rest of the analysis was completed without correlating the Module 2 Quiz scores with the pre-test scores.

Module 2 Quiz
At the conclusion of the instructional unit the students completed a twenty-question quiz composed of eleven kinematics questions, from the Hestene's test, and nine questions from a pool assessing additional content. Testing for homogeneity using Levene's test, F(5,55)=0.512, p > 0.05, and the kertosis, 0.007 and skew, 0.307, of the data illustrated that the data set fit the general assumption.

No significant difference was found between the three groups (CON, M = 10.95, SD = 3.154; SBC, M = 12.50, SD = 3.400; SAC, M = 11.14, SD = 3.044), F(2,50) = 1.516, p > 0.05. Analysis for temporal influences were not significant, F(1,60) < 0.001, p > 0.05, eliminating concerns associated with the delay between course offerings, and, as with the pre-test, the gender distribution did not lend itself to analysis

Hestenes Questions
A more detailed analysis of subject performance on the impact of the simNewton activities was comprised of the eleven questions from the Hestene's set. Reliability tests were run on each item and two were found to be questionable, reducing the reliability coefficient alpha. Removing these two questions produced increased Alpha from 0.703 to 0.780. Testing for homogeneity on the reduced set of nine questions using Levene's test, F(2,58)=0.969, p >.05, the kertosis, -0.745 and skew, 0.004, of the data confirmed that the data fit the general assumption.

Group Mean Score SD Lab Activity
CON 4.57 2.675 Traditional Acceleration Lab (written) F(2.61) = 2.304,
SBC 6.17 2.007 SimNewton lab before content was given p > 0.05
SAC 4.95 2.400 SimNewton simulation after content was given

No significant difference was found between the three groups (CON, M = 4.57, SD = 2.675; SBC, M = 6.17, SD = 2.007; SAC, M = 4.95, SD = 2.400), F(2,61) = 2.304, p > .05. Analysis for temporal influences were not significant, F(1,60) < .001, p > .05, eliminating concerns associated with the delay between course offerings, and, as with the pre-test, the gender distribution did not lend itself to analysis.

Effect Size and Power
In all three analyses the effect size increased from low in the pre-test, h2 < .01, to medium in the Module 2 Quiz, h2 = 0.027, to high in the Hestene's questions, h2 = 0.074. While the sample sizes were well within those determined using Pearson-Hartley tables (for a power of 0.80 a minimum sample size of 15 for each treatment was estimated), the resulting power determined by SPSS for each case were all < 0.50.

Discussion
The no significant difference results in both the Module quiz and reinforced by the examining only the results on the questions from the Hestene's test imply that the web based simulations using simNewton were at least as effective as the traditional lab exercise in supporting students' development of a conceptual understanding of kinematics. Further, the results do not support the findings of Brant, et al. that introducing the lab simulation prior to content improved the development of the student's conceptual framework.

The results of this study are limited by two factors. The results may only be generalized to female students due to the large imbalance in gender (approximately 90% in each treatment group). In all cases the subjects were non-traditional community college students, thus limiting generalization to no more than the online community college population.

While the study demonstrated that the simulation laboratory exercise was at least as effective as the traditional physical lab exercise, there was hope that the simulation would prove more effective. The conflicting power results indicate that a more extensive sample may provide an answer as to whether the simulations are more effective or not. A study needs to be carefully designed to gather a balanced population in terms of gender, and it would be of interest to see if the age of the participants is a factor in the outcomes or effectiveness of the simulations. There is anecdotal evidence that younger students of the "video game" generation are more likely to respond and interact with simulation tools. A more extensive study accounting for these additional factors should extend the educational community's understanding of the impact that newer, web-based, interactive tools may play in developing conceptual understanding of physical systems.

Andre, T (1996) Mission Newton! And Thinker Tools: Using Prior Simulations to Promote Learning about Motion. Iowa State University.
Brant, G., Hooper, E., & Sugrue, B. (1991). Which comes first the simulation or the lecture? Journal of Educational Computing Research, 7(4), 469-481.
Choi, B.-S., & Gennaro, E. (1987). The effectiveness of using computer simulated experiments on junior high students' understanding of the volume displacement concept. Journal of Research in Science Teaching, 24(6), 539-552.
Clement, J. (1982). Students' preconceptions in introductory mechanics. American Journal of Physics, 50(1), 66-71.
Frederiksen, J. R., & White, B. Y. (1998). Teaching and Learning Generic Modeling and Reasoning Skills. Interactive Learning Environments, 5, 33-51.
Frederiksen, J. R., White, B. Y., & Gutwill, J. (1999). Dynamic mental models in learning science: the importance of constructing derivational linkages among models. Journal of Research in Science Teaching.
Halloun, I. A., & Hestenes, D. (1985). Common sense concepts about physics. American Journal of Physics, 53(11).
Heller, J.I., & Reif, F (1984), Cognition and Instruction, 1, 177.
Hestenes, D., Wells, M., & Swackhamer, G. (1992). Force concept inventory. The Physics Teacher, 30, 141-158.
Jonassen, D. H., Peck, K. L., & Wilson, B. G. (1999). Learning with Technology: A Constructivists Perspective. Upper Saddle River, NJ: Merrill, Prentice Hall.
Jonassen, D. H. (2000). Revisiting Activity Theory as a Framework for Designing Student-Centered Learning Environments. In D. H. Jonassen & S. M. Land (Eds.), Theoretical Foundations of Learning Environments (pp. 89 - 121). Mahwah, NJ: Lawrence Erlbaum Associates.
Land, S. M., & Hannafin, M. J. (2000). Student-Centered Learning Environments. In D. H. Jonassen & S. M. Land (Eds.), Theoretical Foundations of Learning Environments (pp. 1 - 19). Mahwah, NJ: Lawrence Erlbaum Associates.
Larkin, J. (1980). Cognition of Learning Physics. American Journal of Physics, 49(6), 534-541.
Leonard, W. H. (1989). Ten years of research on investigative laboratory instruction strategies. Journal of College Science Teaching, 18(5), 304 - 306.
Liu, Han-Chin (2002). Investigating the use of ThinkerTools to promote learning of Newton's laws of motion among middle school students. Iowa State University.
Lunetta, V. N., & Hofstein, A. (1981). Simulations in science education. Science Education, 3, 243-252.
McDermott, L. C. (1984). Research on conceptual understanding in mechanics. Physics Today, 24-32.
McDermott, L. C., & Redish, E. F. (1999). Resource Letter on Physics Education Research.: United States Department of Education.
Mestre, J., & Touger, J. (1989). Cognitive research-what's in it for physics teachers. The Physics Teacher, 447-456.
Mestre, J. (1991). Learning and instruction in pre-college physical science. Physics Today, 56-62.
Posner, G. J., & Strike, K. A. (1989). The conseptual ecology of physics learning. Paper presented at the Annual Meeting of the American Educational Research Association, San Francisco, CA.
Redish, E. F. (1994). The implications of cognitive studies for teaching physics. American Journal of Physics, 62(6), 796-803.
Redish, E. F., Saul, J. M., & Steinberg, R. N. (1997). On the effectiveness of active-engagement microcomputer-based laboratories. American Journal of Physics, 65, 45-54.
Redish, E. F. (1998). Millikan Award Lecture: building a science of teaching physics. American Journal of Physics, 67, 562-573.
Schommer, M. (1993). Comparisons of beliefs about the nature of knowledge and learning among postsecondary students., University of Illinois, Urbana-Champagne, IL.
Shchank, R. C., Berman, T. R., & Macpherson, K. A. (1999). Learning by Doing. In C. M. Reigeluth (Ed.), Instructional-Design Theories and Models: a New Paradigm of Instructional Theory (Vol. II, pp. 161 - 182). Mahwah, NJ: Lawrence Erlbaum Associates.
Thomas, R., & Hooper, E. (1991). Simulations: an opportunity we are missing. Journal of Research on Computing in Education, 23(4), 497 - 514.
Trowbridge, D. E., & McDermott, L. C. (1980). Investigation of student understanding of the concept of velocity in one dimension. American Journal of Physics, 48(12), 1020-1028.
Weller, H. G. (1996). Assessing the impact of computer-based learning in science. Journal of Research on Computing in Education, 28(4), 461-486.
White, B. Y., Frederiksen, J. R., & Spoehr, K. T. (1993). Conceptual models for understanding the behavior of electrical circuits. In M. Caillot (Ed.), Learning Electricity and Electronics with Advanced Educational Technology. New York, NY: Springer Verlag.
White, B. Y. (1993). Intermediate Causal Models: A Missing Link for Successful Science Education. In R. Glaser (Ed.), Advances in Instructional Psychology (Vol. 4, pp. 177-251). Hillsdale, NJ: Lawrence Erlbaum Associates.
White, B. Y., Shimoda, T. A., & Frederiksen, J. R. (1999). Enabling students to construct theories of collaborative inquiry and reflective learning: computer support for metacognitive development. International Journal of Artificial Intelligence in Education, 10(2).
White, B. Y., & Frederiksen, J. R. (in press). Technological tools and instructional approaches for making scientific inquirt accessible to all. In M. Jacobsen & R. Kozma (Eds.), Learning the sciences of the 21st century: theory, research, and the design of advanced technology learning environments. Mahwah, NJ: Erlbaum.
White, B. Y., & Scwarz, C. V. (in press). Alternative approaches to using modeling and simulation tools for teaching science. In N. Roberts & W. Feurzeig & B. Hunter (Eds.), Computer modeling and simulation in science education. New York, NY: Springer-Verlag.