In April of 1995, the UCF Virtual Communities Project Team conducted three weeks of experimental evaluation of the ExploreNet system at Hungerford Elementary School (HES). The simulated "world" we used was based on an incident described in the autobiography of Zora Neale Hurston.

The experiments were successful in identifying strong and weak points of the ExploreNet system and in providing substantial guidance for the next cycle of lesson design. However, a number of technical failures prevented the completion of the formal quantitative evaluation of learned vocabulary and story content.

The teachers and staff of Hungerford Elementary School were extraordinarily helpful hosts and collaborators. The children were enthusiastic, creative and worked hard to achieve the goals of the project. The experiments were, in fact, the high point of the 1994/95 UCF Virtual Communities Project. We intend to continue and expand the Project's relationship with the school and the Eatonville community.

Contents

1. Goals of the Virtual Communities Project
2. Goals of the Hungerford Experiments
3. Experimental Design
4. Measurements of Performance
5. Description of the Trials
6. Lessons Learned
7. Future Plans
8. Acknowledgements
9. Bibliography

 Appendix A: The Egg Quest Story

1. Goals of the Virtual Communities Project

Educational technology has a long history of high expectations and modest actual impact. With the exception of audio-visual aids such as overhead projectors and VCRs, most ed-tech is marginalized - used by a small proportion of the students in a school, for a small percentage of their school time. The reasons are complex, but the effect is that K-12 students and teachers are the only significant working population in America still using 19th-century tools and methods.

An alternative model to traditional "delivery-of-information" teaching is being advocated in the form of the "student as worker" model (Sizer 92.) If schools can be reorganized so that students take responsibility for the completion of coherent, meaningful projects similar to adult work, it is believed that the quality and quantity of their learning will substantially increase. How can technology assist in this transformation?

An essential part of productivity technology in 1995 is the Internet. But meaningful tasks and working environments must be designed for students; simply "surfing the net" is enjoyable but too unstructured to guarantee desirable results. In 1991 the authors began to design and construct a software system called ExploreNet, intended to provide an exciting work-space for student projects that could be shared between schools, or between grades within a school.

The basic idea is to create simple but richly interactive "virtual worlds" using ordinary, affordable PC equipment networked together. Instead of (or in addition to) traditional writing assignments, arts and handicrafts, ExploreNet students would create interactive adventures - a kind of simulated theme park. They would populate these simulated worlds by acting as "cast members", and invite other students to tour their handiwork. By taking on responsibility for others' learning, the students would themselves learn more. This constructionist model (Harel 91) has proven quite effective in other contexts.

The overall educational theory of ExploreNet is described in (Moshell 95a.) These experiments were carried out using Version 3.1 of the ExploreNet software, which is described in (Hughes 95.)

2. Goals of the Hungerford Experiments

Before we can consider having students create world-simulations, we needed to understand more about the educational use of these simulations. In particular we needed to see if students of various ages could control the simulations, and if they could effectively serve as cast members for other students. That is, could they learn a story line well enough to act out its roles, so that "guests" - new students who don't yet know the story - could interact with the simulation and learn the desired lessons?

Hungerford Elementary School in Eatonville, Florida was chosen as the site for this initial experiment for several reasons. Hungerford was one of Orange County's first prototype elementary schools for technology; every classroom has three to six networked 286 computers. The teachers, staff and students have several years of experience with computing and are familiar and comfortable with keyboards and mice. Hungerford is also a feeder to Maitland Middle School which is Orange County's first prototype middle school for technology, and the main developmental site for ExploreNet. Hungerford is sited in the historic town of Eatonville, which is the first municipality in the United States which was created by, and has always been administered by African-Americans. The school's population is 95% Black.

As the FoxFire project (Wigginton 72) has shown, students can be strongly motivated by curricular connections to local culture. Eatonville was the home of Zora Neale Hurston, an internationally-renowned African-American folklorist and novelist. Her autobiography provided rich material for the development of interactive stories. Leaders of the Zora Neale Hurston Festival in Eatonville were willing to assist in developing such a body of material, and the school's principal Mr. Leroy Fillmore was willing for the school to participate.

In the fall of 1994, UCF students and faculty developed the first ExploreNet worlds around several mythical talking animals - an alligator, a dog, an owl - taken from African-American folk tales in Hurston's book Mules and Men (Hurston 78). One of these worlds was field-tested in the January 1995 Zora Festival. In these trials, no cast members were used; we simply observed the children's ability to control the characters, and made corrections to the software when problems were noticed.

It was apparent that children as young as 4 years old could control ExploreNet, and that they loved the animated characters. We therefore began to design an experiment.

3. Experimental Design

The details of this experiments' design are found in (Moshell 95b.) What follows here is a brief summary.

Concepts. The essential idea was to work with one class each of third, fourth and fifth grade students as study groups, and to use an equal number of classes and students as control groups. In the study group, teams of four students would first experience the world as guests - that is, they would learn how to use the system and then play the game. The cast members for this first experience would be university students.

Once the Hungerford students had been guests in an initial 45 minute experience, they would know the tricks and secrets of winning the game. They would then serve as cast members for another group of students. The experimenters would observe how the overall interaction moved forward, which parts of the software worked well or poorly, and would administer several pre- and post-tests to the students to see if they learned from the experience.

Variables. The dependent variables were to be measures of vocabulary acquisition and the complexity and richness of stories written after the experience, by the participants. The independent variables were to be:

1. Computer versus traditional treatment
2. Age of cast members and age of guests
3. Shared screens vs. one student per screen

The Game. The actual game to be played was called The Egg Quest, and was based on a very simple story in Dust Tracks on a Road, Hurston's autobiography (Hurston 70.) She reported that she and her friends would sometimes go to the barn, steal chickens' eggs, build a fire, boil the eggs and eat them. We elaborated on this idea by making the chickens to be wily opponents, protecting their eggs.

The students had to discover that they could put corn out for the chickens, who would then hop off their nests. Once the students (using animated human characters named Zora, Zeke, Andy and Anna) grabbed the eggs, they had to find firewood and a pot, make a fire and cook the eggs. The story had been pilot-tested at Page School on University Boulevard in March, and was well accepted by the students. At Hungerford, most groups of guests succeeded in capturing the eggs within the 45 minute activity period; about half the groups managed to cook them.

Initial Training. When students first encountered the system, they would be assigned the role of one of the human figures - Anna and Andy (Team A), or Zora and Zeke (Team B). All appeared to be pre-teen children. The characters' names were posted on signs atop the computers in use. Anna and Andy shared a single computer. Zora and Zeke each had their own computer. Anna and Andy were guided by a mentor character named Jake the Dog; Zora and Zeke were guided by Wise the Owl. Initially the system is locked so that the guests cannot leave this first scene. We had discovered during the first week's trials that new students will leave the scene by stumbling across exits, and that training then becomes very difficult.

The adult trainers would first lead one student (typically Zora) through the operation of the mouse, and show them how to move the character and make it talk. Then all four guests would try out these actions. After the mentor character determined that they were well trained, the mentor character selected a menu item which released the guests to the next scene.

In the second scene, the adult guides would show the guests how to pick up, put down and use 'props' - objects such as lanterns, firewood or eggs. This scene also contained the lantern for starting a fire, and either firewood or a cooking pot, which would be needed later in the scenario. (Zora and Zeke saw only the firewood, and Anna and Andy saw only the pot.) Once these basic skills were mastered, the mentor would reset the game to the initial state and release the guests to travel freely. The guests could then travel from starting place, through the prop-introduction scene (and pick up anything that might be useful) and then move to the final scene which was the barn. The EggQuest world is laid out as shown in the following diagram.

Figure 1: Scenes of The Egg Quest

The two groups would thus be trained separately, and would only meet one another at the barn scene. It was expected that they would exchange information or collaborate, since one group would have the wood (or knowledge of it) whereas the other would have the cooking pot.

Intended Schedule. We had intended to run five groups of four students per day, and to use four of the groups for statistical analysis. (The fifth group was scheduled so we could give all students in the class a chance to participate.) Each session was to run for 45 minutes. The plan was to run groups with UCF cast members and HES guests on Tuesdays, and HES cast members and guests on Thursdays. One week each was allocated for third, fourth and fifth grades; though some students from third and fifth would participate in each others' trials. However the numerous logistical problems prevented this pattern's completion, as described in Section 5 below.

4. Measures of Performance

In consultation with the Hungerford teachers it was decided to measure vocabulary acquisition and story comprehension. The teachers provided a working vocabulary list of around 200 words, from which we selected ten to incorporate into the story. An elaborate mechanism was designed for scoring pre and post-tests for vocabulary and for evaluating stories written by the students.

Mr. Frank Haynie, the Hungerford Curriculum Resource Teacher (CRT), was the project's principal liaison with the school. Mr. Haynie agreed to teach a lesson to the control groups, based on the same story line as the ExploreNet experiment. He initially read a story to the groups (see Appendix A for the story) and then had the students draw pictures to illustrate the story. The same pre- and post-tests were given to these groups as to those in the study groups.

As will be seen, a number of factors ultimately made this quantitative evaluation plan unworkable. However, the overall experimental design was carried out and a number of useful lessons were learned. They are reported in the next section.

5. Description of the Trials

On Tuesday, 11 April 1995, we arrived at Hungerford Elementary with a van full of tables and computers. We set up the equipment in a straight line across the east side of Ms. Tribue's fourth grade classroom, in the configuration shown in the following figure.

Figure 2: Computer Setup for Hungerford Trials

The UCF party for each trial consisted of Michael Moshell[1], Daniel Tan[2] and Homer Whittaker[3], plus one or two graduate students. Bruce Caulkins and Hector Morelos among the grad students came to several sessions; the others came to one or two sessions. Thus there were typically four or five outsiders, plus five or six computers and two cafeteria tables, being added to the already crowded classrooms. The teachers were remarkably tolerant of this disruption.

During these trials which involved four or eight class members, the classroom's normal activity continued. The teachers were quite effective in maintaining the class' focus, in part because the computers' screens were facing away from the class. Also, the teachers did not generally attempt to do whole-class instruction during the trials. Hungerford has considerable experience with students working in groups, in part because of their exposure to IBM's TLC Integrated Learning System. Using TLC with three to six computers for 24 to 30 students necessarily means that the class will be working in small groups on several topics at once.

We began to work with groups of four students, using UCF personnel as cast members. However, within an hour it was apparent that the software had severe reliability problems, and so we cancelled that week's trials and removed all the equipment. Throughout these trials we set up and removed all the equipment each day, as there wasn't room in the classrooms to leave the gear in place when we weren't there running experiments.

On Tuesday 18 April we resumed with Ms. Tribue's class. Due to logistical difficulties, we only had two full sessions with UCF cast and HES guests. The students were assigned to groups in alphabetical order by last names, based on who was present thast day. On Thursday of that week, we began by running a third UCF/HES group. Then we used the three groups who had already been guests, as cast members to provide the same experience to three more groups.

On Tuesday 25 April, we moved to Ms. Davis' fifth grade class. Because of the loss of the first week, it was necessary to work with both third and fifth graders in this setting. On Tuesday we provided the UCF/HES experience to three fifth grade groups and one third grade group. On Thursday, we had the following groups: HES5/HES3 (i. e. Hungerford fifth graders as cast members for Hungerford third graders); U/HES3, HES5/HES3, HES5/HES5, and two UCF/HES3 groups.

Pre-tests. Mr. Haynie had arranged for pre-tests for vocabulary to be administered to the students. However, due to communications errors the fourth grade was not pre-tested.

Post-tests. In each classroom we had a circular table set up adjacent to the computer table, and we would move the group that had just finished the computer trials to this table to administer the post-tests, while the next group of students used the computers. It proved very difficult to proctor this activity. The children were quite crowded; the table could comfortably accomodate at most five students; with eight, they were packed in like eggs. There was nowhere else available to take the children for testing at this time.

The excitement level was quite high, having just (in most cases) successfully completed the Egg Quest. The students merrily assisted one another, hassled one another and generally acted like happy elementary students would, under the circumstances. The teacher (and in the case of Ms. Tribue's fourth grade her intern), attempted to assist the project's staff in proctoring the post-tests, but they had the rest of the class to supervise. It proved impossible to assure that each student was doing his or her own work.

In addition, a number of the sessions were terminated early due to software failures. Some of the others were immediately before lunch time and thus no post-test was possible. At the end of the experiments it was apparent that the data was incomplete and that such data as we had, was not gathered in a meaningful manner. No analysis was attempted.

Summary. In all, 24 (six groups of four) fourth graders experienced the Egg Quest. Three of these groups served as both cast and guests; the other three groups were guests. In the fifth grade, 20 students (five groups of four) had experience as guests; of these, three groups had cast experience. From the third grade we had six groups of four students as guests, and none as cast members.

The total number of students who had the Egg Quest experience at least once was 68; of these, 24 also served as cast members.

6. Lessons Learned

Each project staff member provided written feedback, which was then integrated into this report. The lessons learned are sorted into the following categories.

6.1 Logistics
6.2 Software and User Interface Design
6.3 The Cast/Guest Model
6.4 Experimental Design
6.5 Story Design

6.1 Logistics

Location. There was initially some consideration given to the idea of using a work-room in the media center and bringing small groups of students out of the class. The teachers requested that the activity be done in class, so as to minimize the chance of kids getting lost in the halls and to keep everyone under their supervision.

The main problem with this plan was the lack of space and time for supervised post-testing.

Recommendation: Plan better for post-testing and allow for providing the necessary isolation of students from one another.

Initial Training. Out of each four trainees in a group, one would consistently look away from the demo, sometimes to concentrate on the screen in front of them. Perhaps the screen should be blank until time to play. Some students seemed to be pressured to "keep up with the group" even though they hadn't mastered basic skills. Students without the basic skills seldom observed the advanced ones to learn their techniques.

Recommendation: Provide several training tasks, so that students who learn skills faster can continue to train rather than becoming impatient to move on.

6.2 Software and User Interface Design

The following observations have already resulted in substantial changes to ExploreNet 3.2.

Two Button Mouse. The division of mouse functions across two buttons was a serious impediment to the rapid learnability of the system. We have now moved to a one-button method for the next version.

Lack of sound. Without sound effects or speech there is no easy way to attract the user's attention to a guide character. However in classrooms, the use of speaker based sound would distract adjacent learners, so earphones seem essential for this feature. Text to speech technology is being investigated for ExploreNet 4.

Difficulties with Props. Some students found it difficult to pick up objects. The menu structure for putting down objects was two layers deep. We have improved the menu structure for picking up objects so that now a single click on a prop provides a list of its affordances (actions.) The option to pick up the object is the first choice on this menu list.

Text Input. The means of entering text was clumsy. A large input window was provided wherever the cursor was located. In the new version a small fixed window is provided at the bottom of the screen for text.

Typing. Lack of typing and spelling skills seems to be an obstacle to keeping up a sufficient pace. Composing text messages was time consuming and leaves other players wondering what that person is doing. With the older students (fifth graders) this was less of a problem and they often got engaged in composing elaborate messages. However, their fellow-travelers didn't know what was happening and so would get bored, or abandon the earnest writers.

Recommendations: Construct stories for younger children that do not rely very much on textual input from the students. If the encouragement of reading is a goal, then make sure that the "speech" of your cast members is highly relevant to the story line.

Make it possible for other guests to know that one of their member is engaged in writing a message. Consider showing the message character by character as it is composed, to increase the immediacy of the communication.

Identity. The second most frequent question was "Who did that?" There was a strong desire to identify the other characters' players. Students often wanted to direct messages to specific other players. Anonymity may play only a limited role in these systems, or may only be useful in voting and brainstorming.

The identity of the speaker adds context to the message. Therefore, the benefits or disadvantages of anonymity may be dependent on the specific group of learners. Do they have common goals? In today's classroom, students and teachers often appear to have very different goals, with traditional learning being a high priority goal only for the teacher.

We have not yet constructed scenarios that are sufficiently engaging (and it may be impossible in principle) to submerge the strong interpersonal dynamics acting between school-age children. It may be that only when we are really telecommunicating with strangers will this "who's that" phenomenon become less of a distraction.

Recommendation: Think about ways of using this strong desire to communicate with the actor rather than the character, to our advantage rather than as an obstacle to engagement. Perhaps let the kids design their own on-screen characters (e.g. from an avatar-builder kit) and use their own names.

6.3 The Cast/Guest Model.

The Mentor Character. Animal guides were not much used as sources of information. The tendency was to seek help from adults rather than animal guides. This may be because speech is easier than typing and reading. Some students said that they want to be able to talk to other characters more easily. Also - adults, conveniently close by, represent authority in traditional classroom culture. Sometimes when students asked for help and it wasn't available, they would simply wait rather than "talk" to the animal guides or observe others.

The most frequent question was "how do I ..." and was almost always asked of an adult mentor; occasionally of a student sitting alongside; and almost never of an on-screen character.

Recommendation: make the adult guides less available so as to emphasize the importance of working with the on-line mentor characters. Explore text-to-speech as a possible output medium to make the on-screen mentors more engaging.

Guides had a tendency to play the game rather than guiding the guests. Sometimes they would give commands rather than hints to the guests. On some occasions the guides were reluctant to help the guests. This may be due to insufficient guide training and experience.

Recommendation: Experiment with middle-school children to see if they can stay focused on the role of guide, rather than playing the game. Fourth and fifth graders from our study population had difficulty doing this. Study ways of better training the cast members. The ones in this experiment had only one guest-experience (of 45 minutes duration) and about 10 minutes of training in cast member responsibilities.

Cast Members' Knowledge of Guests' Actions. Often, guides are in the dark as to what guests were doing. They would either observe inactivity, or actions that didn't seem to have anything to do with the game. Sometimes the apparent inactivity was due to the guest's slowly typing out a message or seeing assistance from an adult.

Recommendations: Guides should be able to see what guests are doing; what menus or windows are open, etc. They should be able to ask if the guest needs help, at any time - even if the guest is in the midst of some operation. This might need to be a voice channel because guests are so engaged in their work that they ignore screen activities they aren't generating.

6.4 Experimental Design

Pilot Tests. Our ambitious experimental design, which consumed many weeks of design effort from several people, was rendered meaningless by logistical problems that would have been discovered if we had tried more one-day pilot trials. The pilot at Page School was very helpful but was not structured like the Hungerford trials and so did not teach the essential lessons.

Recommendation: Pilot small. Think hard, then pilot again. Pilot your controls as well as your main tests. Get order-of-magnitude ideas about what kinds of variables you will encounter.

Example: Frank Haynie's control students uniformly wrote longer, richer stories than the computer students did. How, precisely, did he instruct them? We didn't allow time to observe that, and we need to be able to understand the reason for such a variation before we try to measure anything based on the stories.

Another example: The problem of not having adequate space and time for the post-tests, which rendered the results invalid, would have been detected with pilot trials.

Conducting the Experiments. Projects such as this one, without external funding, rely crucially on volunteer labor and the willing participation of faculty members "on speculation" - in the hopes that significant results and substantial grants will follow. We received substantial input and excellent advice from Art Cross, Donna Walker-Knight and Cindy Woodley of the College of Education; Kathy Sheridan and Ed Rinalducci from Psychology, and Kay Stanney from Industrial Engineering.

These behavioral scientists and educators who contributed to the experimental design served essentially as consultants to the project; the actual experiments were carried out by computer scientists. A consequence of this was that many mistakes were made which would have perhaps been avoided if professional experimentalists were on-site during the planning and design as well as the execution of the experiment.

However, the computer scientists got what they came for - extensive testing of the software and of the overall didactic model of role-playing games. The attempt at multi-disciplinary collaboration was educational for all.

Recommendations. We now know that we need to write proposals and secure funding for enough of the time of our behavioral colleagues that they can afford to be intimately involved in both the design and conduct of the experiments. Any faculty member with a three-course assigned teaching load is not going to have significant amounts of "speculative time" to provide active un-funded support to projects such as this one.

6.5 Story Design

Reactions. The elementary students liked the story, both in written and computer form. The "quest" format seemed natural to them, and we had almost no one who was not actively engaged in trying to solve the central puzzles of the story line. The story was of about the right level of complexity for a 20 to 30 minute encounter for fourth graders, after an initial 10 to 20 minutes of learning how the user interface worked.

Fifth graders were capable of completing the story in less time, but the particular fifth grade class with which we were interacting was in a rather rambunctious mood due to the imminent end of the semester. This contributed enough "noise" to the process that their actual times were not much different from the fourth or third graders.

Recommendation: Make stories somewhat longer and more complex than the Egg Quest if it is targeted at fifth through eighth graders. The complexity is about right for fourth and below. The literacy level required may make the textual portions of the game useless for very young players.

Collaboration. The principal failure of the story from its creators' point of view, was that it did not induce collaborative behavior. The students within a two-person group were sitting next to one another and thus could collaborate by talking out loud. Collaboration was expected to be needed when the eggs were uncovered, as one character was to drop the corn at some distance from the chickens, while another stood ready to get the eggs. In fact, usually the corn-carrier would drop it quite close to the chickens. When a chicken would uncover the eggs, all the characters would make a mad dash for the eggs. The resulting pile-up would make it difficult for anyone to pick up the eggs.

Once someone had succeeded, it was impossible for the others to tell who had the eggs. The classroom would frequently echo with cries of "Who got the eggs!?"

In a similar vein, rather than each team having half of the firewood/pot setup, usually the most aggressive player of the four would have discovered that props are useful and would have circulated throughout the world, picking up everything. The rest of the quest would then devolve into a lot of debate about what should be done, with one player holding all the assets. Those kids usually looked rather smug. When this would happen, one of the four guests would often drop out and start star-gazing or daydreaming or verbally interacting with the rest of the class.

Recommendation: If you intend to force collaborative behavior you need to build in tasks which CANNOT be accomplished by a single player who is much more aggressive than the others.

7. Future Plans

(What other kind are there?)

Training. Ultimately we hope to have the system's interaction fine-tuned enough so that we can have the mentor character provide initial training, rather than requiring a cast of over-the-shoulder tutors. This will be a very challenging task, and is intimately related to designing a viable model for teacher involvement in the introductory stages of ExploreNet use.

Dependent variables. The choice of dependent variables is a major issue. In this experiment the choice was driven by teacher preferences; this naturally led to the use of traditional content-oriented measures. In the future we expect to try harder to define measures that relate to skills learned.

Technology. ExploreNet 3.2 is already much more user-friendly than 3.1, on the basis of lessons learned at Hungerford. We will continue to add features such as sound, and to build more worlds for the exploration of the cast/guest relationship. In the fall of 1995 we move into a new phase, as Maitland Middle and Coral Springs Middle Schools begin world-building experiments.

Future experiments at Hungerford. In March, the project team assisted Hungerford Elementary in writing a Superintendent's Grant proposal for seventeen Pentium computers so as to continue the use of Explorenet in the fall of 1995, in conjunction with Maitland Middle School. In addition, the project in collaboration with the Coalition of Essential Schools, Orange County Schools and schools in California, Boston and New York, has submitted a $13.6 million Department of Education Challenge Grant proposal. If this is funded, Hungerford will receive substantial financial support for technology and curricular innovation.

In any case we intend to continue to work with Hungerford and to participate in the 1996 Hurston Festival.

8. Acknowledgements

The project sincerely thanks Orange County Public Schools and Mr. Leroy Fillmore, Hungerford's principal, for the opportunity to conduct these experiments in the school. Mr. Frank Haynie, the Curriculum Resource Teacher, was a superb collaborator and made all things possible. And the three teachers -

Ms. Chambers, third grade
Ms. Davis, fourth grade
Ms. Tribue, fifth grade

- were the heart and soul of the experiment. These ladies are heroes, and our children are in good hands.

We also appreciate the support of the Association to Preserve Historic Eatonville and its Director, Ms. N. Y. Nathiri. She participated in the design of several ExploreNet scenarios, arranged for the provision of a portable classroom for the Zora Festival experiments, and provided insights into Hurston's work.

Among the project's volunteers, Mark Kilby and Lamar Harrell provided exceptionally thorough written feedback and reflections on the experiment; their notes form much of Section 6.

9. Bibliography

Harel, Idit and Papert, Seymour. Constructionism. Ablex Publishing Corporation, Norwood, NJ. 1991.

Hughes, Charles E. and Moshell, J. Michael. "Shared Virtual Worlds for Education: The ExploreNet Experiment." submitted to ACM Multimedia Journal, June 95.

Hurston, Zora Neale. Mules and Men. Indiana University Press, Bloomington, Indiana. 1978.

Hurston, Zora Neale. Dust Tracks on a Road. U. of Illinois Press, Champagn-Urbana IL. 1970.

Moshell, J. Michael and Hughes, Charles. The Educational Model for the Virtual Academy. Document VC95.14. Computer Science Dept., University of Central Florida, Orlando, FL. April 1995 (95a)

Moshell, J. Michael; Stanney, Kay; Rinalducci, Ed; Cross, Art; Sheridan, Kathy and Frederick, Terry. Evaluation Plan for the Virtual Communities Project. Document VC95.11. Computer Science Dept., University of Central Florida, Orlando, FL. April 1995 (95b)

Sizer, Theodore R. Horace's School: Redesigning the American High School. Houghton Mifflin Co., Boston, MA. 1992.

Wigginton, Eliot (ed.) The Foxfire Book. Anchor Press, Garden City, NY. 1972.