Willis, J. (1995). A recursive, reflective instructional design model based on constructivist-interpretivist theory. Educational Technology, 35 (6), 5-23.

The epistemology underlying a constructivist-interpretivist approach to ID does not permit the creation of a single ID model that represents this theoretical perspective. There are many possible models, each with potential strengths and weaknesses depending on the context of use. One model that has emerged from work at NASA's Johnson Space Center and the Center for information Technology in Education at the University of Houston is the R2D2 Model: Recursive, Reflective Design and Development. Like many behavioral or objectivist-rational ID models, the R2D2 model also has several components - Define, Design and Develop, and Disseminate. They have the same names as components in the Four D model of Thiagaraian, Semmel, and Semmel (1974). However, in the 4D model the components are generally treated as sequential, linear pleases. Most traditional ID models are based on a sequential set of stages or phases, which can be conveniently represented by flowcharts. In the R2 D2 model the four components are focal points. The R2D2 model is non-linear. Designers do not necessarily focus on the Define component first. The model is also recursive or iterative as well as reflective. Aspects of the Design and Develop component, for example, may be the focus of attention several times during the instructional design process.

That is because the ID team (which may be large or small) is expected to actively reflect on and analyze work to date and regularly revise and rework both the material being developed and the models that underlie its development. In many ways the recursive-reflective aspect of the model is similar to Turkle's (1984) "soft mastery" approach to using a computer, while traditional linear ID models are examples of her "hard mastery" approach. The P2D2 model is represented graphically in Figure 1.

The figure is an adaptation of Escher's "three tribars" (Ernst, 1994). Escher became interested in "impossible tribars" after reading an article by R. Penrose in the British Journal of Psychology, (Vol. 49, part 1, February, 1958; cited in Ernst, 1994). An "impossible tribar" is a triangle that has three 90 degree angles. Escher linked three tribars together in a drawing that was the inspiration for Figure 1. (Escher's abstract, three-tribar figure was the basis for one of his most famous lithographs, Waterfall, in which water drops off a waterfall and then "unquestionably flows continually downward, and at the same time recedes from us. All of a sudden, the furthest and lowest point turns out to be identical with the highest and nearest point; therefore, the water is able to fall once again and keep the wheel turning" (Ernst, 1994, p. 88). In Figure 1 the three tribars make up one large tribar with each side representing one of the focal points of the model: Define, Design and Develop, and Disseminate. The reflective and recursive nature of the R2D instructional design model is suggested by the lack of a visual center or focal point. Instead, the figure invites the eye to "travel" around it over and over again just as the water in Escher's Waterfall continuously moves around the "impossible" circuit.

There are many differences between traditional ID models and the R2D2 model presented here. Two, however, are so broad and general that they influence the entire instructional design process. Instead of progressing linearly from stage to stage as tasks are completed, the R2D2 model is recursive. That is, the same issues or questions may be addressed many times throughout the process. Solutions, decisions, and alternatives gradually emerge over the course of the project.

A second broad principle of the model is Schon's (1987) reflective approach to professional practice. Whereas objectivist ID models tend to put great emphasis on work both before (developing objectives) and after (summative evaluation) the actual work of creating instructional materials, the R2D2 model puts the emphasis on the creative process itself. Schon's concepts of reflection-in-action and reflection-in-action have much to offer the instructional designer. They move us away from the objective, technical approach that involves matching Current Conditions to pre-made solution templates-the "technical-rational" approach. Instead, the emphasis is on the current context-a complex and changing situation that must be understood in order to make good professional decisions. Reflective practice involves gaining greater insight into both the context and the influence of your work on that context. "The reflective practitioner assumes a dual stance, being, on one hand, the actor in a drama and, on the other hand, the critic who sits in the audience watching and analyzing the entire performance" (Osterman & Kottkamp, 1993, p.19).

Reflective practice is a modern expression of an established tradition:

Figure 1. A graphical representation of the R2D2 Instructional Design (ID) Model. The model has three focal points (Define, Design and Develop, and Disseminate). The nature of the graphic shown above, which has no obvious beginning or ending and constructs an "impossible world" perspective, represents the two Rs of the R2D2 ID model: Recursion and Reflection.

In the R2D2 model, the instructional design process is viewed as a learning process in which learning leads to improvements in the instructional material under development. Reflection is a primary avenue of learning and is situated within what Dewey (1938) called "the problematic," in this case, the process of creating instructional materials. It is active, authentic, social, and collaborative because it occurs during the process of development and involves a team of participants who cooperate to make decisions.

Reflection is also a recursive or cyclic process. "Experiential learning theory maintains further that learning is a dialectic and cyclical process consisting of four stages: experience, observation and reflection, abstract reconceptualization, and experimentation" (Osterman & Kottkamp, 1993, p. 20).

In the R2D2 model, much more than in objectivist ID models, most of the activity is in the context of creating the instructional material. There is much less emphasis on both preliminary activities and work that comes after the material has been developed. Reflection and Recursion occur primarily in the Design and Development aspects of instructional design-thus the name R2D2.

To illustrate how the R2D2 model looks in practice, this section will describe its application in a joint NASA/University of Houston project that involved the creation of computer-based instructional materials for adult literacy programs. The instructional material, called LiteraCity, is based on authentic instruction and simulation models. Students select a simulated "friend" who needs to find a job. They help the friend search for a job, read help wanted ads, select promising openings, and fill out employment applications. The program, which is distributed on CD-ROM, was developed using the R2D2 ID model. (in this section, traditional ID models are represented by the general format of the Four D model mentioned earlier, but many other models use essentially the same framework.)

In the traditional ID model, this task includes the preliminary steps required to develop any type of instruction. The purpose of Definition subtasks is the identification and definition of instructional requirements. The five traditional subtasks in this stage include (1) front-end analysis, (2) learner analysis, (3) task analysis, (4) concept analysis, and (5) specifying instructional objectives.

Front-End Analysis. The initial step in developing instructional material is an evaluation of the need. This process generally involves a specification of the need and an evaluation of existing materials to determine if needs are already being met. When the R2D2 team took over this project, a traditional front-end analysis had already been conducted. It indicated that some computer-supported instructional materials were available for adult literacy programs, but almost all of them were drill and practice or tutorial packages designed from a distinctly behavioral perspective. No materials based on constructivist approaches were found.

Learner Analysis. This step traditionally involves developing a clear understanding of the target students. In this case, the primary users were adult literacy students. We treated learner analysis as an on-going process rather than as a preliminary phase of development. We worked with a number of students at the major adult literacy program in Houston, but the format was not a traditional learner analysis that involves assessment of learner skills. Instead, the students became part of the development team. We asked for input on interesting themes for the simulation, asked about topics that should be in the simulation, and received feedback on visual "look and feel" issues as well as on instructional strategies the students preferred. Instructors at the literacy program also gave input on what their students needed, what they preferred, and ways the simulation could be made more interesting and relevant.

Task and Concept Analysis. After the need has been established and relevant student preferences, characteristics, and needs defined, the next step in a traditional ID model is to define the major skills to be acquired by the students and to analyze those skills. In traditional behaviorally-based projects, the focus is on breaking the content down into subunits or components that can be taught separately. In reading, this approach leads to a discrete skills model of instruction that targets things like short vowel sounds and consonant digraphs. The result of this model in adult literacy programs is often relatively boring exercises that attempt to remediate weak underlying skills on the assumption that when students develop those cognitive "muscles," their reading will improve because reading is really a collection of skills. This model has not served adult literacy well and it may be one reason for the high dropout rate in many programs as well as the lack of enthusiasm shown by many participants who truly want to learn to read. In addition, many adults who do not have a functional level of reading ability have already completed and/or failed at many such exercises during their school years and have significant negative associations with them.

Using the R2D2 ID model, we approached task and concept analysis from a somewhat different perspective: authentic instruction and assessment. The primary goal was to deliver instruction and provide assessment directly related to authentic tasks an adult in this culture completes by reading. Those tasks are the focal point of the project, and both instruction and assessment focuses on them. Thus, during instruction, the task of the reader is to accomplish a common task such as finding and applying for a job. Assessment focuses on the reader's success in accomplishing the task. That is not to say that discrete skills are ignored. It simply means that they are framed within the context of how they relate to a particular task that is real or authentic for the student.

When task and concept analysis are considered within the framework of authentic assessment and instruction, the analysis becomes less a job to be done once and for all at the beginning of the development process and more a continuous aspect of ID. For the project described in this article, the task was "Finding a Friend a job," which involved selecting appropriate jobs, obtaining and completing the application, and submitting the application to the potential employer. The instructional product was built around the task. Actual instruction and help is available to the student when, and only when, the student requests it. Students select help and instruction when they need it to complete the authentic task at hand.

Specifying Instructional Objectives. This is normally the final step in the first stage of a traditional ID model. It involves converting the results of the analysis of tasks and concepts into a set of objectives. They in turn provide the basis for designing the instructional package and developing evaluation and assessment strategies. This task also changes when viewed from the perspective of the R2D2 model. In behavioral theory, specific instructional objectives are needed to guide the design process. In the R2D2 model, specific objectives evolve naturally from the process of design and development. In the LiteraCity project, for example, the general goal of developing instructional materials for adult literacy programs was set early. However, specific behavioral objectives were never actually written down and used to guide instructional development. instead, students and teachers discussed the type of tasks they wanted to learn to do. That led to the selection, by the students, of "finding a job" as the focus of the first package. (Another simulation, begun by a previous ID team, that involved renting a movie at a video store was abandoned because students felt finding a job was far more important to them.) The authentic task selected by students and literacy teachers thus determined the focus of the instructional material, and the requirements of that task determined the type of reading activities that were included in the simulation. Thus, in the R2D2 model, it is not important to write specific objectives at the beginning of a project. It may even be impossible to do that because the specific focus and direction of the ID project may not be well understood. What is important from the beginning is to involve end users, in this case, teachers and students, in the entire design process. This approach has much in common with the participatory design approach to workplace redesign and computer systems development that is better known in Scandinavia than in North America (Schuler & Namioka, 1993).

The focus of the Design Stage in a traditional ID model is the creation of a design for the instructional material. Two subtasks of design in most ID models, materials selection and format selection, are usually completed along with a third, selection of a development environment. Two other traditional design subtasks, initial design and selection of an evaluation strategy, round out the work to be completed during the Design Stage.

The Design Stage work is traditionally completed before the Development Stage, when the instructional material is actually created. The Design Stage is usually where some members of the team think about the package and create paper representations of it such as flowcharts or storyboards. The initial design prototype, for example, may be a storyboard. The next stage, the Development Stage, is where the material is actually created-first as a prototype; then, as formative evaluations of the prototype are obtained, revisions are made. However, if the data reported by Wedman and Tessmer (1992) are any indication, many instructional designers do not regularly conduct formative evaluations. Their prototype may thus quickly become the final version that is ready for summative evaluation.

The separation of design from development is a potential source of significant difficulties. Heyer (1987) describes what is, hopefully, a worst case scenario with regard to laserdisc and CD-ROM development:

Heyer paints a depressing picture of some of the problems that can occur when components of design and development are separated. He emphasizes the separation of design and development tasks as well as development team members, but similar points could be made about design and development work that does not remain in constant contact with end users.

For many reasons, the R2D2 model combines the traditional Design and Development Stages into one focus point. The Design and Development focus of the R2D2 model has four components: media and format selection, selection of a development environment, product design and development, and evaluation strategy. Each of these components is discussed below, but as you will see, the boundaries between each component are not well structured or strong.

Media and Format Selection. The media we selected was interactive multimedia and the format selected was simulation of authentic tasks. These choices emerged from discussions with teachers and students as well as from reviews of the literature and exploration of emerging multimedia instructional materials.

Selection of a Development Environment. Although there are many theories and models for selecting development environments, the fact of the matter is probably that development teams usually select authoring environments and support software they already know how to use. That was the case with this project. The core of the development environment was Authorware Professional on the Macintosh (APM), supplemented by sound and graphics software, including Photoshop, SoundEdit, and Canvas. There were, however, other reasons for selecting APM. It has a number of advantages over standard programming languages such as C. Perhaps the most important advantage when using an ID model that calls for extensive collaboration is that it is relatively easy to learn. Computer-literate team members can be "up and running" in Authorware in less than a week, while estimates of the time it takes someone to become a proficient C programmer are as high as two to three years. This is important in the R2D2 model, because many members of the team may be involved in "programming." In traditional ID models, team members often take a specialist's role. For example, a graphic artist may create illustrations, an instructional designer produces sequences or frames of instruction, and a content expert either writes, selects, or approves the content included in the package.

In the R2D2 model, as used in this project, the team was composed of "specialists" who were primarily responsible for different aspects of the program such as graphics or sound. One member of the team served as the designated programmer. The roles were not, however, watertight compartments. Instead, almost all the team members, with the exception of literacy students and teachers who did not have the time, did some work on all aspects of development. The content specialist, for example, wrote and revised some of the Authorware code and redesigned some of the icons. The sound specialist wrote some of the script for the simulation and designed several of the help and instructional strategies. APM's ease of use made it possible for every team member to program.

The interactive nature of APM also made design and development recursion possible. Without a development environment like APM, reflection and recursion would be less useful, perhaps useless, because it is too difficult to make changes and adjustments to the program once it is "set in code."

Much of the traditional emphasis on creating precise instructional objectives and producing detailed flowcharts and storyboards is a reflection of the time, effort, and expense involved in making major revisions to materials produced in traditional programming languages. However, the R2D2 model encourages experimentation and exploration of alternatives. APM supports that approach. For example, instead of storyboarding an instructional sequence and then turning it over to the programmer to be coded, the LiteraCity team could quickly answer many "what if" questions. What if we change the location of the icons, what if we add an explanatory screen between two existing screens, what if we change the layers a reader moves through to get help with an unknown word or phrase? In APM it is easy to modify and revise the instructional material and then run that segment of the program to see how it will appear to students. The same feature makes it very easy to get student and teacher input and suggestions on different components as the program is evolving. For example, the team created four different "look and feel" designs and asked sixty adult literacy students to select the one they preferred. The look and feel most members of the development team predicted the students would prefer was actually one of the least preferred. The same group of students also listened to eight musical styles and selected the one they liked best - which was again not the one most of the team predicted.

Heyer (1987) has also proposed that tasks of design and development be combined. He sees three major advantages to this approach. Risk is reduced because fundamental problems are discovered early. Because design and programming occurs at the same time, there is less likelihood that the team will commit to a design that cannot be executed. A second advantage is the encouragement of experimentation and revision which will enhance the quality of the final product. Finally, the process of design and development becomes understandable and "each person knows exactly what the others are doing."

Product Design and Development. An educational product is as much an artistic creation as it is a technical product. We are all aware of technically correct educational materials, manuals, and documentation that numb the mind and discourage the user. The LiteraCity project aimed to create a product that appealed to the user; one that had the snap and crackle that comes from effective use of graphics and sound, and an interesting premise creatively executed. A traditional, top-down, linear model of development is unlikely to produce such a product. What is needed is a development model that is the equivalent to the "management by walking around" movement in corporate leadership. That is, the designers and developers need to be able to "walk around" the program and try out possible alternative designs and arrangements with regular and ongoing feedback from potential users and consumers. And, with an appropriate development environment, changes, once made, can be run immediately to view the effect from the perspective of a student who will be using the package. Being able to see the effects of a change immediately avoids the designer-to-programmer-to-software-to-designer cycle that has doomed so many instructional development projects. It also allows developers to "play with" many different alternatives in much the same way a musician can play with changes in a musical composition. This non-linear, real-time approach supports improvization, which leads to exciting designs. Eric Holsinger (1992), in his article titled "Nonlinear systems enhance video editing," makes the same point about electronic video editing. "Unlike standard video-editing systems, an editor can quickly cut and paste sections of the presentation together in any order. A user can save and compare different edits of the presentation easily, whereas starting a new version on a traditional off-line system requires you to start the edit from scratch each time."

This approach was implemented in the LiteraCity project. Changes, even major changes, were relatively easy to make. Major changes were, in fact, made throughout the ID process. There was never a time when a design element was "locked in" and could not be changed. instead of being determined before the development phase began, the final format, look and feel, and content "emerged" across the development process.

Evaluation Strategy. The evaluation strategy used in the R2D2 model is actually an integral part of the design and development process. It emphasizes formative evaluation and puts relatively little emphasis on summative evaluation. Formative evaluation is emphasized because the purpose of the project is to create the best possible instructional material, and summative evaluation contributes nothing to that it simply tells us if we have been successful. Formative evaluations, on the other hand, provide feedback the team can use to improve and enhance the instructional package. The types used in this project were:

            2. Expert Appraisal

As noted earlier, students and experts frequently looked at components of the package before even a rough version of the full package was available. Students as well as experts evaluated many components including alternative graphic designs, background music, and screen layouts. The information obtained from component feedback was used to guide further design and development. Then, when a prototype of the package (with a "single path" as explained below) was available, both experts and students provided detailed critiques of many versions. This approach to evaluation has much in common with Eisner's (1979) connoisseurship model and with literary and art criticism.

A design and development approach that is recursive and allows for changes and revisions in "real time" calls for a collaborative team with a wide range of skills and expertise. It also calls for a team that can tolerate almost continuous change, adjustments, and fine tuning as well as the frustration that is probably inevitable when some work is abandoned or replaced by new material. The major drawback to traditional linear models of instructional development may be that they eliminate many opportunities for experimentation, fine-tuning, and artistic enhancement. The major drawback to recursive models may be that everything is so fluid that team members have difficulty seeing progress and feeling positively about their accomplishments to date. There are, however, ways to add some semblance of structure to the work and to provide the team with some internal indicators of progress. In the LiteraCity project, we worked concurrently on three aspects of the general format and design:

1. Surface Design-screen layout, typography, language, graphics, illustrations, sound.

2. Interface Design-look and feel, user interaction, help, support, navigation, metaphors.

3. Scenario-sequence of simulation options/ choices, results.

We approached these three components by defining one path through the simulation that included every type of interaction, every type of help, and every type of outcome and feedback. We then completed all the work on that path so that the simulation ran from start to finish as planned-if one particular path was selected. Initially, we used "placeholder" graphics, photographs, sound, and instructional sequences. At that point we were able to shift from component feedback to package critique. Literacy students and teachers, as well as other experts, were then able to critique this path, and revisions were made based on 'he feedback from those critiques. Gradually, placeholder materials were replaced with more polished final components. First, the appropriate graphics, sound, and images were created and added to the single path. Then, after teachers and literacy students as well as experts in relevant areas evaluated the material, a pattern and format emerged from work on the single path and it was extended to the other paths through the program.

In a traditional ID model, the dissemination stage has at least four components: summative evaluation, final packaging, diffusion, and adoption. With the exception of summative evaluation, dissemination in the R2D2 model is quite similar to that in behavioral ID models. Final packaging involves creating and producing any necessary print materials (such as student guides and teacher manuals), creating a master of the instructional materials (print, software on disk, CD-ROM, or laserdisc), and producing the package in quantity. Advertising, training sessions, workshops, creation of a user support system, and updates of successful packages round out the dissemination focus.

On the subject of summative evaluation, however, there are significant differences in behavioral ID models, a summative evaluation is an objective assessment of the effectiveness of the instructional package. It usually involves objective tests of one sort or another, and the results of the summative evaluation are often treated as "proof" the program works. In the behavioral, empirical-objective tradition, successful summative evaluations are sometimes even treated as "proof" other instructional packages with similar characteristics are also valid. In the R2D2 model, summative evaluation becomes a minor, and not always necessary, component. There are several reasons for this lowered status. First, since the instructional package developed is only one aspect of successful instruction, it is difficult to generalize from the summative evaluation to other contexts. With different teachers, different students, the likelihood other teachers will use the materials in somewhat different ways, and a different school and community context, a successful summative evaluation is certainly no guarantee that the material will work the same way in another context. Also, traditional summative evaluation models tend to rely on objective measures because of the objective emphasis in behavioral approaches and because they lend themselves to the task of evaluating the type of instruction behavioral approaches promote. Constructivist approaches often encourage individual goal setting by students and advocate diverse learning activities among a group of students, even when they are "studying the same thing." Objective tests are not always a good fit with this type of learning. Projects, diaries, activity logs, reflective journals, reports, the assessments of mentors, and portfolios are often more appropriate, but they do not always lend themselves to traditional summative evaluation procedures.

In the LiteraCity project, we treated the final expert appraisal as part of the summative evaluation. Also part of the "summative" evaluation were data from the final student tryouts and the students' performance on the authentic tasks built into the instruction. We concluded that students who completed the simulation enjoyed it and could read want ads and complete job applications-the goal that emerged for LiteraCity during development. Ironically, when one of the NASA executives (now retired) saw the program in operation, he was quite impressed, but kept asking about grade level changes in reading. We explained that the ability to read want ads drawn from the types of papers these students would use to find a job was an "authentic" means of assessment that directly measured a relevant academic skill, but he remained adamant that the only way to "really" know if the program worked was to test students on a standardized reading test. Like the NASA executive, behavioral educational technologists are not likely to find our approach to assessment satisfactory. Again, however, I would argue that while objective measures of academic progress can be useful, they can also be limiting, misleading, and downright unhelpful. The creation of "good" instruction may be more like the creation of art than building a bridge that does not fall down. If that is the case, we may need to follow Eisner's (1979) lead and look closely at evaluation and assessment procedures in the arts and humanities.

Alternative answers to two core issues-the role of language and the definition of truth-lead us to an instructional design model that is less rigid, less authoritarian, less confident of decisions, and more than a little fuzzy. In a recursive, non-linear model, many decisions are made over and over, and developers begin the process of instructional design without a crisp, clear definition of where they are headed. This is an inevitable consequence of an antifoundational approach. If we cannot be confident that we know what is "out there" in reality, then we must always be open to revision and change. However, recursive approaches, like linear models, can be taken to extremes that are both frustrating and nonproductive. Whistler, for example, had great difficulty deciding when a painting was truly "finished." On several occasions, he actually went to the homes of patrons who had purchased paintings and made small changes to them. A team that can tolerate a process in which many-things remain fluid and changeable, however, will gain opportunities for fine-tuning and artistic enhancement.

The collapse of the empiricist's objective-subjective dichotomy has many implications for instructional design. Many fear that the loss of an objective foundation for research and practice puts us straight on the road to hell-in this case, a wimpy relativism that treats every opinion and theory equally, since there is no way to make objective decisions. Rorty's brand of interpretivism goes beyond the simple assertion that everything is subjective. Since we cannot escape our own backgrounds and experiences, he believes we must practice, design, conduct research, and develop theories within that framework-it is our only option. We cannot work in a vacuum and we cannot shed our background.

We need not remain, however, totally within our own group. Members of the behavioral community of instructional designers, for example, who develop multimedia tutorial and drill and practice software, need not remain completely ignorant of the work of other groups, such as constructivists, who create software to support anchored instruction. As Rorty (1991) put it:

Rorty's democratic pragmatism thus accepts the subjective nature of both research and professional practice, but he argues that we can reduce the danger of what he calls "ethnocentrism" by encouraging free and open discussion within our own group and by making a special effort to seek out and understand the truths of other groups.

Bagdonis, A., & Salisbury, D. (1994, April). Development and validation of models in instructional design. Educational Technology, 34(4), 26-32.

Bednar, A., Cunningham, D., Duffy, T., & Perry, J. (1992). Theory into practice: How do we link? In T. Duffy & D. Jonassen (Eds.), Constructivism and the technology of instruction (pp. 17-35). Hillsdale, NJ: Lawrence Erlbaum.

Cognition and Technology Group at Vanderbilt. (1990). Anchored instruction and its relationship to situated cognition. Educational Researcher, 19(6), 2-10.

Cognition and Technology Group at Vanderbilt. (1991). Technology and the design of generative learning environments. Educational Technology, 31(5), 34-40.

Cognition and Technology Group at Vanderbilt. (1992). The jasper experiment: An exploration of issues in learning and instructional design. Educational Technology Research and Development, 40(l), 65-80.

Criswell, E. (1989). The design of computer-based instruction. New York: Macmillan.

Dewey, J. (1938). Logic: The theory of inquiry. New York: Holt, Rinehart & Winston.

Diamond, C. (Ed.). (1976). Wittgenstein's lectures on the Foundations of mathematics: Cambridge 1939. Chicago: University of Chicago Press. (Note: Diamond's book is based on the class notes of four students who attended the lectures.)

Dick, W. (1987). A history of instructional design and its impact on educational psychology. in 1. A. Glover & R. R. Ronning (Eds.), Historical foundations of educational psychology. New York: Plenum.

Dick, W. (1995, July/August). Instructional design and creativity: A response to the critics. Educational Technology, 35( 4), 5-1 1.

Dick, W., & Carey, L. (1985). The systematic design of instruction. Glenview, IL: Scott, Foresman.

Eisner, E. (1979). The educational imagination: On the design and evaluation of school programs. New York: Macmillan.

Ernst, B. (1994). The magic mirror of M. C Escher. New York: Taschen America.

Fullan, M. (1993). Change forces: Probing the depths of educational reform. Bristol, PA: The Falmer Press.

Gagne, R., & Briggs, L. (1979). Principles of instructional design (2nd ed.). New York: Holt, Rinehart, & Winston.

Giacomi, J., Mosher, A., & Swenton-Wall, P. (1993). Ethnographic field methods and their relation to design. In D. Schuler & A. Namioka (Eds.), Participatory design: Principles and practices (pp. 123-155). Hillsdale, NJ: Lawrence Erlbaum.

Gropper, G. (1987). A lesson based on a behavioral approach to instructional design. In C. Reigeluth (Ed.), Instructional theories in action (pp. 45-112). Hillsdale, NJ: Lawrence Erlbaum.

Heyer, M. (1987). The creative challenge of CD ROM. In S. Lambert & S. Ropiequet (Eds.), The new papyrus (pp. 347-358). Bellevue, WA: Microsoft Press.

Holsinger, E. (1992, October 14). Nonlinear systems enhance video editing. MacWeek, 10, 29.

Howe, K. (1988). Against the quantitative-qualitative incompatibility theses (or, dogmas die hard). Educational Researcher, 17(8), 10-16.

Kemp, J. (1985). The instructional design process. New York: Harper and Row.

Kuklick, B. (Ed) (1981). William James: Pragmatism. Indianapolis: Hackett Publishing Company.

Lieber, J. (1991). An invitation to cognitive science. Oxford, UK: Basil Blackwell.

McAlpine, L. (1992). Highlighting formative evaluation: An instructional design model derived from practice. Performance and Instruction, 31(10), 16-18.

Merrill, M. (1988). Don't bother me with instructional design - I'm busy programming! Suggestions for more effective educational software. Computers in Human Behavior, 4(l), 37-52.

Osterman, K., & Kottkamp, R. (1993). Reflective practice for educators. Newbury Park, CA: Corwin Press.

Reigeluth, C. (Ed.) (1983). Instructional-design theories and models: An overview of their current status. Hillsdale, NJ-. Lawrence Erlbaum.

Rorty, R. (1979). Philosophy and the mirror of nature. Princeton, NJ: Princeton University Press.

Rorty, R. (1982). Consequences of pragmatism. Minneapolis: University of Minnesota Press.

Rorty, R. (1991). Objectivity, relativism, and truth: Philosophical papers volume 1. Cambridge, UK: Cambridge University Press.

Schon, D. (1987). Educating the reflective practitioner. San Francisco: Jossey-Bass.

Schuler, D., & Namioka, A. (1993). Participatory design: Principles and practices. Hillsdale, NJ: Lawrence Erlbaum.

Schulte, J. (1992). Wittgenstein: An introduction. Albany, NY: State University of New York Press.

Senge, P. (1990). The fifth discipline. New York: Doubleday.

Simon, H. (1991). Models of my life. New York: BasicBooks.

Skinner, B. F. (1984). The shame of American education. American Psychologist, 39, 947-954.

Smith, J. (1989). The nature of social and educational inquiry: Empiricism versus interpretation. Norwood, NJ: Ablex.

Smith J. (1993). After the demise of empiricism: The problem of judging social and educational inquiry. Norwood, NJ: Ablex.

Spiro, R., Feltovich, P., Jacobson, M., & Coulson, R. (1992a). Cognitive flexibility, constructivism, and hypertext: Random access instruction for advanced knowledge acquisition in ill-structured domains. In T. Duffy & D. Jonassen (Eds.), Constructivism and the technology of instruction (pp. 57-76). Hillsdale, NJ: Lawrence Erlbaum. Originally in Educational Technology, May 1991.

Spiro, R., Feltovich, P., Jacobson, M., & Coulson, R. (1992b). Knowledge representation, content specification, and the development of skill in situation specific knowledge assembly: Some constructivist issues as they relate to cognitive flexibility theory and hypertext. In T. Duffy & D. Jonassen (Eds.), Constructivism and the technology of instruction (pp. 121-128). Hillsdale, NJ: Lawrence Erlbaum. Originally in Educational Technology, September 1991.

Turkle, S. (1984). The second self. New York: Simon & Schuster.

Thiagarajan, S., Semmel, D., & Semmel, M. (1974). Instructional development for training teachers of exceptional children: A sourcebook. Reston, VA: Council for Exceptional Children.

von Neuman, J. (1951). The general and logical theory of automata. In L. A. Jefress (Ed.), Cerebral mechanisms in behavior. New York: Wiley. (Cited in Simon, 1991.)

Vygotsky, L. (1978). Mind in society. Cambridge, MA: Harvard University Press.

Wedman, J., & Tessmer, M. (1992, April). Instructional designers' decisions and priorities: A layers of necessity study. Paper presented at the annual conference of the National Society for Performance Improvement, Miami.

Willis, J., Stephens, L. & Matthew, K. (in press). Computers, reading, and language arts: A guide. Boston: Allyn & Bacon.

Winograd, T., & Flores, F. (1987). Understanding computers and cognition. Reading, MA: Addison-Wesley.

Wittgenstein, L. (1958). Philosophical investigations (3rd ed). (Trans. G. Anscombe). New York: Macmillan.

Wittgenstein, L. (1965). The blue and brown books. New York: Harper and Row.

Wright, C., & Conroy, C. (1988). Preparing CBI Print-Based Support Material: An Information/instructional Design Perspective. Distance Education, 9(l), 84-94.

[part 1]     [part 2]     [back to module]

Willis, J. (1995). A recursive, reflective instructional design model based on constructivist-interpretivist theory. Educational Technology, 35 (6), 5-23.

© 1995, Educational Technology
Permission being sought.