This article is the outgrowth of a meeting of biology teachers, researchers, and software developers at Beloit College (see sidebar). Our purpose was to discuss how our efforts, until now separate, might be joined to create a new undergraduate biology course. While we all use computers and are involved with the development of instructional software, computers, per se, are not the issue here. Rather, this article is about how computers and a new pedagogy can increase the importance of content-based, problem-solving learning in biology. This paper proposes that computers can facilitate student laboratory research, and that such research is central to a deep understanding of how scientific knowledge is created, modified, and employed. Further, domain-specific, problem-solving is a powerful vehicle for teaching the content of the domain along with its working procedures.

Our experience is in teaching biology, but we feel that the thesis of this article applies across science learning, and we have seen evidence of similar thinking in the pages of this journal in fields ranging from anthropology to psychology. We hope our ideas challenge you to examine areas in your own discipline where information is being memorized by students when problem-solving could convey both information and fundamental ideas about the operation of the discipline.

The pedagogy underlying this paper is motivated by a fundamental respect for the learner. We believe that learning experiences should be humane, that students deserve to be treated as fellow learners, not infantilized. This conviction comes from self-reflection and a recognition that we have learned most, and appreciated the learning most, when our abilities as learners were respected in the learning setting. We hope that throughout this paper you can put yourself in the place of the student and reflect on how it is that you would most like to learn.

The Problem In General Biology Education

We believe that an unrealistic approach to science learning has metastisized throughout our courses. Textbooks are dry and static, labs are cookbook, lectures push information to students who have become junkies for mythical "scientific facts." In reality there is no such thing as a fact unsullied by theory (or prejudice or hunch). The store of information in biology is doubling several times a decade, and the theories that shape our knowledge are constantly changing. The details of what we teach in an introductory course are of necessity incomplete, if not wrong, even as our students sweat to memorize them. If we continue to walk the rutted path, we will defraud them of the real understanding of science for which they are paying us. What could endure longest from an introductory course is an understanding of, and some practice with, the way biologists pursue their craft. Our courses must help students understand how biologists: perceive the world; pose questions; pursue the problems from those questions; and, ultimately, persuade others of the value of their solutions.

We are motivated by the failure of current science instruction to provide students with important integrative thinking skills. Stewart and Dale (1981) studied high school students solving problems in Mendelian genetics. They found the students frequently got "right" answers without understanding their methods or the genetic processes implicit in those methods. Collins (1986) studied a variety of genetics "experts" solving similar problems and observed that a deep understanding of genetics underpinned their problem-solving. Significantly, she notes that they resorted to fundamental material and mechanical models when they were studying a problem. That is, the geneticists in her study would employ a chromosomal model of meiosis and fertilization, then, reasoning by first principles, they would work to understand seemingly aberrant or unusual patterns in their data. Another significant difference between the students and experts is the rigor of their solution proofs. Collins's experts went through three stages that tested their solutions: a confirmatory cross, a definitive cross, and a disconfirmatory cross. Students, on the other hand, were not systematic in double checking and ruling out alternative hypotheses. Such a distinction between novice and expert is not learned in a single class, rather it results from different expectations about the nature of "doing" science. 

Herbert Simon (1982) has called the type of integrative thinking demonstrated by Collins's expert problem-solvers a "science of the artificial." By this he means a formalizable body of knowledge about the process of creating man-made (as opposed to nature-made) artifacts. Simon's concerns are broad-ranging, from engineering (design of objects) to economics (design of business plans) to political science (design of social organizations). When we apply Simon's notions of design to science, the hypothesis becomes the artifact to be created. To develop a hypothesis, the problem-solver "constructs meaning" from data. This view of science learning departs from students' ideas about "finding the answer." Finding answers implies an algorithmic approach will "solve" a problem (plug the numbers into the equation). Construction of meaning is not algorithmic and requires different expectations about science and the intellectual commitment it requires. 

Metaphors like "design of a hypothesis" or "construction of meaning" may not, at first, seem appropriate to science education. Certainly, such activities are far from present methodologies. We use these terms because of the similarities that we find in the process of science that are common to the process of learning/working in other creative domains. A recent issue of Machine Mediated Learning was devoted to examples in which the metaphor of design was used as a vehicle to teach content-specific problem-solving across the undergraduate curriculum. Examples were presented in English composition, computer programming, architecture, and science. In each case the authors used computers to create design-like situations that incorporated a common process of drafting, testing, and revising. Students were guided in the creation of artifacts: an essay, a building, a computer algorithm, a testable hypothesis. 

Discipline-Specific Learning Through Problem-Solving

The authors in Machine Mediated Learning were not interested in teaching abstract notions about problems and their solutions. Rather, problem-solving was a vehicle for studying the deep content issues in their discipline. Adolph (1978) describes scientific research as "a mainspring of [self-] education, as a means of self-formation and self-renewal." He intertwines the idea of knowledge about problem solutions with knowledge about the system under investigation. We concur with Adolph's emphasis on the importance of self-education through research and feel that a good knowledge of biology involves experiencing first-hand the production and application of scientific knowledge. 

Non-scientists can also benefit from a science education that sheds light on the production and application of hypotheses and the nature of scientific knowledge. John Moore (1986) describes the virtues of such an education with an example of students learning the difference between creationism and evolutionary science as ways of explaining. There is no correct answer as to what one should believe on this subject, but the source of evidence one chooses and the way one works with that evidence to defend conclusions illustrate different ways of knowing. Similarly, we argue that computer-aided classroom research can teach how knowledge arises from explanations based on evidence. 

"This is all well and good," you might argue, "but I need to cover so much material in a term that I can't get past the presentation of information to cover these more sophisticated aspects; besides, students need information to think with before they can think." There is some truth to that argument, but we are not proposing that problem-solving learning come at the expense of lecture and reading. On the contrary, there is a powerful synergism between the two kinds of learning experiences. Rather, the concern of this article is to improve the laboratory. Labs have atrophied because they are expensive, difficult, and time consuming and, often, aren't well connected with the other materials in the course. But, laboratories and laboratory-based learning have a time-honored role in science education. The Master's and Ph.D. degrees are clear evidence that both content and methodologies are learned in laboratory research. The goal of this paper is to illustrate how the richness and lessons of the research laboratory can be available to a freshman in a general biology course. 

The Practicum

Central to the research component of graduate science training is the practice it provides in "doing" science. The reason problem-solving is so important in learning about science is that this is where so many of the blatant biases of culture and ideology enter. Prejudices based on race, gender, sexual preference, ethnicity, class, and religion are exceptionally well documented in the annals of science. However, other biases such as reductionism, teleology, anthropomorphism and anthropocentricism, speciesism, and confirmation-bias all too often creep into our formulation of problems as well as into our strategies for testing hypotheses and persuading peers. If hidden assumptions are ignored during problem-posing, it is all too difficult to eradicate them at the problem-solving and persuasion stages of scientific practice. 

Second, problems do not come pre-posed to scientists. Frequently, a problem's solution is much less difficult if it has been well posed. Problems may have a historical thread woven into them, and the perceptions employed to pose problems may in themselves be theory-laden; nonetheless, students need to understand clearly that they could stand in the lab or field forever, and no textbook-stated problems would come to them out of thin air. Students can only begin to appreciate the tremendous agenda-setting issues in problem-posing if they are encouraged to pose problems themselves. 

For example, Dorothy Buerk (1982) at Ithaca College frequently begins her math talks with the following problem: "If one person could paint a house in four days, then how many days would it take four people?" She said that you need to understand the context; if the problem is at the back of a chapter on "one equation with one unknown algebra" or is on a general aptitude exam, then answer: "One day." However, if you are examining problem-posing, then you can play Brown and Walter's (1983) "What if...?" game. What if the painters belong to a union? Only have two brushes? Enjoy playing foursome bridge? Are male? It rains? Are paid by the hour? Such games rapidly illustrate to students that problem-posing is inherently laden with linguistic conventions of taste, implicit assumptions, and cultural perception. 

The most basic aspect of textbook and tutorial problems is that they ask questions of the students. Even simple computerized simulations can go beyond this by requiring the student to ask the questions, and this is a critical feature of a genuine scientific education. When students begin asking questions, they will rapidly see that their problem-posing and problem-solving activities are inseparable (Jungck, 1985), and that drawing warranted inferences depends upon knowing the assumptions that were made at the posing stage. 

Donald Schšn (1987) has studied the role of the practicum in professional education. His study focused on architectural design, master classes in music, and clinical psychology. In each case, the core of the training involved realistic practice with the tools and problems of the profession: designing buildings, playing music, psychoanalyzing patients. We call problem-solving in science education "realistic" when it captures the open-ended essence of science as it is practiced: problems must be both posed and solved by the problem-solver. In contrast, most general biology courses are taught with "unrealistic" problems that: come pre-posed, have unique answers arrived at unambiguously, and are checked for correctness by an authority. This paper illustrates how computers help create a practicum in the general biology laboratory, 

Schšn's teacher is a "coach" who alternates between demonstrating for the student as they work jointly on a problem and critiquing the student's solo performance. For example, consider the problem of learning to swim. It cannot be learned by expert swimmers lecturing about swimming or swimmers. Rather, learner and coach get in the water and practice. The learner must feel himself swimming in order to understand what swimming is. This is very different from the undergraduate biology course where the instructor expounds and the student memorizes. The laboratories described below are settings in which the teacher's role is as coach, demonstrating techniques, critiquing work in progress, and working as a colleague rather than as an authority on problems in which neither student nor teacher knows the answer. 

Computers And The Laboratory Practicum

Early in the experimentation with educational computing, Thomas Dwyer (1974) recognized that "deep technology is of little value without a deep view of education." Previous science education software failed, in our view, because it lacked a view of education sufficiently deep for its technology. For computers to be pedagogically powerful, they must provide students with realistic problem-solving environments. Three kinds of computer software can provide realistic problems: simulated laboratories, professional software tools, and general productivity tools. All three should be integrated into the general biology laboratory because the key to learning in a scientific practicum is the ability to work realistically with data generated in the laboratory. 

For the laboratory practicum to be successful, we believe that students must experience:

• problem-posing, 
• problem-solving, and
• persuasion.

We call these the "three Ps" of scientific practice. For us, the three Ps are as essential to teaching science as the three Rs were to education at the turn of the century, and the key to learning the three Ps is practicum experience. Learning the three Ps will be illustrated with three different software tools. 

Problem-Posing

First, students need experience in posing original and significant (at least to them) scientific problems which are capable of being solved. The term "simulated laboratory" has been used to describe classroom applications of computers that provide realistic simulations of professional laboratories (Peterson and Campbell, 1984). These computer programs allow students free experimentation with apparatus and analysis tools commonly available in a well equipped research laboratory. Strategic Simulations in Mendelian Genetics, (abbreviated GCK for Genetics Construction Kit) by Jungck and Calley (1985) is one such simulated laboratory. It presents problem-posing at its simplest, Figure 1. The program begins by preparing a random "field collection" of an organism with different phenotypic traits. In the figure, the sample organism is a hypothetical fruit fly. Some specimens have white eyes, others red. Some have bent wings, others stubby, still others straight. The program provides the tools for performing genetic crosses. It also provides statistical spreadsheets for analyzing the outcomes, and a notepad for recording hypotheses. Given these tools, the student's problem is:

• What cross to make first? 
• Which of several traits to pursue? 
• What hypotheses can be drawn from the initial data? 
• Can these be confirmed? 
• When confirming a first hypothesis, does the new data point to a new conclusion not anticipated in the field sample?

From the moment the program begins, the student must be active, posing problems that are soluble with the tools provided. There is no tutorial introduction hinting at tasks or outcomes. When frustrated, or when satisfied they have a solution, students can "cheat" and see the genetic mechanism(s) that the computer used to generate the problem. An important feature of GCK is the "research" mode. In this mode cheating is disabled, as is real life, and the computer's "answer" to the problem is not retrievable by either student or teacher. Now, without the ability to refer to an authority, teacher and student must collaborate to arrive at a solution with which they are satisfied. 

One of the members of the our group (see sidebar), Stewart, has used an Apple II version of GCK with high school biology students, mostly ninth and tenth graders. The significance of GCK for these students is that it allows them, as they solve genetics problems, to experience the important activities of posing their own problems and determining when they have a justifiable solution. The program helps create an atmosphere in which students and instructors talk genetics, rather than about genetics in some abstract way. 

The other context in which Stewart has used GCK involves teachers: high school biology teachers, pre-service biology teachers, and university TAs. Invariably, every time GCK is used with these groups some individuals report it has completely changed how they think about Mendelian genetics and, most importantly, how they think about teaching genetics problem-solving. By continually forcing the user to frame problems from the available data and perform crosses that test those hypotheses, GCK forces teachers to reorganize knowledge that they already possess in ways that help solve realistic problems. 

You might contend that "fruit fly laboratories are commonly performed with real flies as part of general college and even high school biology courses. Why should I use a computer to do the same thing?" The strength of GCK is its speed (gestation periods of seconds rather than days), built-in statistical analysis tools, and faithful rendition of the activities that would be undertaken with real insects. GCK has been used in conjunction with live insects to perform pilot studies that help sharpen the plan of laboratory experimentation, avoiding time-consuming dead ends and backtracking with real insects. 

Reviewers experienced with tutorial software that guides students step by step have criticized realistic laboratory simulations like GCK as "too complex." It is true that an open-ended laboratory is more complex than a linear tutorial, but GCK has been successfully used with students as young as grade school. Further, simulated laboratories are necessarily more complex than tutorials because they aim to teach ideas about science that are much deeper and more complex. Are lessons about the nature of scientific knowledge "too complex?" We think not. Rather, as we argued above, memorization of facts trivializes science and deprives students of the opportunity to learn its true nature. 

Problem-Solving

Simulated laboratories are custom pieces of software developed expressly for enhancing laboratory experiences. However, tools used by biology professionals can often provide equally powerful laboratory experiences. For example, the program by Wayne Maddison called MacClade allows the construction of binary classification schemes, Figure 2. Most students consider taxonomy to be an exceptionally dry task; however, if they have to construct a classification which they must support with evidence, then they begin to understand that every classification scheme is a hypothesis. Should the classification be based upon evolutionary ancestry (i.e., parent-progeny relationships)? Should the classification be phenetic (i.e., upon overall measures of similarity)? Should the classification scheme be efficient for identification? These and other questions frequently have contradictory effects in the construction of a classification scheme. 

To help students appreciate the hypothetical nature of classifications and the arbitrariness of some assumptions, we use the time-honored practice of having students collect data of their choice on deciduous trees: leaf shape, size, bark texture, seeds, flowers, arboricity, etc. These indicate to them the pluralism possible in gathering data. Then we have them break down their data into binary categories for ease of analysis rather than making them statistically analyze quantitative characters. If they construct a binary tree from their discrete questions and place their OTUs (operational taxonomic units) on this tree, then you can ask them additional questions such as:

• Is the length of the their tree minimal? 
• Do they have too many parallelisms or reversals? 
• Is their tree optimal for identification? 
• Also, why are their classifications so different from one another if all of them were able to collect data in the same woods, on the same trees, and most of them share a very similar education within the life sciences? 

Persuasion

Research is not part of the scientific community until colleagues have been persuaded. The bottom line is not doing the experiment or analyzing the data; yet, often, student labs stop at these points. We promote the concept that solutions are hypotheses (Collins, 1986) that convince peers by drawing warranted inferences from well collected data. Students need to learn very early in their careers that they haven't done science (no matter how many experiments have been done, how much data collected, how many puzzles solved) until they have both reported their results and convinced their peer group as to the reasonableness of their hypothesis. 

The write-up is the traditional persuasive activity in science laboratories. Unfortunately, practice in scientific writing is not part of many student laboratories. A word processor is an exceptionally useful tool for helping students learn about scientific composition. The speed with which reports can be written and revised with computers allows time for another kind of lesson about scientific writing: it is much more social than tabulations of data in a lab notebook. Biological writing is fraught with the hazards of: teleology, anthropomorphism, circular reasoning, confusing correlation with causation, speciesism, sexism, and racism. 

You might object "there is hardly enough time to correct the simple lab reports students write now. For me to comment on multiple drafts of longer scientific reports will increase work for me by an order of magnitude." To view yourself as responding to all student writing is to participate in the teacher-as-authority paradigm. It is exactly this dramatic increase in work that suggests another paradigm for evaluation must be chosen. Instead, students can be assigned to reproduce and critique each other's experiments, working from the written reports. This converts your task of evaluation into their learning process. Reviewers could be assigned to critique the logical consistency of the paper or develop counter arguments and examples using the same dataset in the same laboratory setting. The carrot for adopting this pedagogy is that the teacher is neither the arbiter of correctness nor the ultimate evaluator. If, instead, you see yourself and your TAs as peers of the students, colleagues more familiar with the literature and accepted arguments, then you can provide input just as you do when reviewing papers for a professional journal. 

Another aspect of science practice to incorporate is multiple authorship. A small team of students will bring more resources to the writing process and be better able to provide useful reviews. For example, at Beloit College, students in Biology publish a "journal." A faculty member serves as the journal's editor. Authors write papers based on their laboratory work, then they must write critical reviews of the papers written by others and revise their own papers in response to critiques. The final products are reproduced in a bound volume. In writing and reviewing for their journal, Beloit's students experience the critical care that is required for successful scientific communication. 

Finally, stressing persuasive communication in conjunction with the construction of meaning in the laboratory has led us to question the structure of the biology textbook. The computer-based hypertext is a new vehicle for presenting textual material. The InterMedia system at Brown University has been used by Peter Heywood to build a biology course that reflects our philosophy of exploring a body of knowledge represented, not in a simulation but in a text. It offers a conversation with primary references, and as students edit and extend the hypertext, a conversation with others. 

Results

Collins has evaluated the problem-solving performance of a variety of genetics students (Ph.D.s with research and teaching experience, high school teachers, university undergraduates, high school, middle school, and even elementary students). Her work has highlighted the important difference between cause-to-effect and effect-to-cause problems. The former are typical of many current textbook problems, the latter describe the thinking that must be employed in the learning experiences described above. Hopkins, et al., (1987) have demonstrated that as little as a five hour, effect-to-cause, problem-solving session in physiology can alter students' perceptions about the kind and strength of interactions between elements of the cardiovascular system. Beginning students were seen to have a very local and anatomic view of the cardiovascular system. Experts know that system-wide interactions are very important to cardiovascular performance. In grappling with explaining how an effect in one place could have been caused somewhere remote, students modified their perceptions of the system itself. These experiences lead to our conviction that problem-posing/problem solving simulations have the potential to revolutionize learning in biology. They present the opportunity for students to do what scientists do: ask questions, formulate tentative hypothesis, design methods to collect data to test these hypotheses, and evaluate the hypotheses in light of the data. 

Conclusion

Since the first attempts at using the mainframe computer for instructional purposes, there have been serious discussions about the way in which the machine should be employed:

• What should it automate because it can? 
• What tedium should it reduce? 
• What roles should (must) be defined for student and human teacher?

Our philosophy requires that the computer provide a framework that allows students and teachers to work as research colleagues on realistic problems. We feel that this philosophy for using computers in teaching biology is sufficiently powerful to justify the cost and effort in employing computing extensively in life science education. 

We do not see a dichotomy between the learning of scientific content and the philosophical nature of science. Rather, these are mutually supportive activities. Presently, the lessons about the nature of science are lost in the laboratory because they are not conducted with tools conducive to teaching those lessons. The computer allows students to construct meaning actively using simulations, as well as professional and productivity tools, thus revitalizing the laboratory as a professional practicum.