Curing Cancer? Patrick Lee's Path to the Reovirus Treatment

Paul Thagard
Philosophy Department
University of Waterloo
Waterloo, Ontario, Canada

Thagard, P. (2002). Curing cancer? Patrick Lee's path to the reovirus treatment. International Studies in Philosophy of Science, 16, 179-193.


This paper provides a historical, philosophical, and psychological analysis of the recent discovery that reoviruses are oncolytic, capable of infecting and destroying many kinds of cancer cells. After describing Patrick Lee's very indirect path to this discovery, I discuss the implications of this case for understanding the nature of scientific discovery, including the economy of research, anomaly recognition, hypothesis formation, and the role of emotion in scientific thinking. Lee's discoveries involved a combination of serendipity, abductive and deductive inference, and emotional cognition.

1. Introduction

People usually think of viruses as dangerous germs that cause diseases like colds, flu, and AIDS. Some viruses even cause cancer, like the herpes virus that produces Kaposi's sarcoma in people with deficient immune systems. Recently, however, medical researchers have found that some kinds of viruses can actually kill cancer cells, and there is much excitement about the prospects of using viruses to treat a wide variety of cancers in people. It is too soon to say whether the new viral treatments will be medically effective, but important scientific discoveries have been made in the past decade about the interactions of viruses and cancer cells.

One of the most intriguing discoveries was made in 1995 by Dr. Patrick Lee, a Canadian virologist. He realized that the reovirus, a medically benign virus that he has investigated for more than two decades, could infect and kill cells with an activated Ras pathway. Ras is a gene crucial for cell growth that, when mutated, can contribute to the uncontrolled cell growth found in cancerous tumors. Most cancer cells have an activated Ras pathway, and the reovirus has been successful in killing in test tubes many kinds of cancer cells, including ones from brain, prostate, breast, ovarian, pancreatic, and colorectal tumors. Moreover, injection of reovirus into mice has produced dramatic elimination of many kinds of tumors, as Lee and his colleagues reported in the journal Science in 1998 (Coffey et al., 1998). A clinical trial is now under way to see whether reovirus injections can also reduce or eliminate cancers in humans, with results expected to be announced in fall, 2001. Because there have been many cancer treatments that worked well with mice but failed with humans, it is unrealistic to expect the reovirus to provide the cure for cancer. But even if the reovirus treatment does not provide the major medical breakthrough that many people are hoping for, its development is an interesting episode in the history of science that illustrates important aspects of scientific discovery.

After providing some background on the history of research on cancer treatments and reoviruses, I will describe the unusual path that led Lee and his colleagues to the discovery that reoviruses can destroy cancer cells. I will then discuss what this case tells us about the nature of scientific discovery, including experimental design, reaction to failed experiments, hypothesis formation, and the role of emotions and aesthetic judgments in scientific thinking,

2. Cancer Treatments

Although many refinements have been made, the main kinds of treatment for cancer have remained the same for several decades: surgery, radiation, and chemotherapy. Cancer was recognized as a disease by the ancient Greeks and Romans, who were aware that it could sometimes be treated by surgically removing cancerous tumors, although cancers were usually fatal (Olson, 1989). After the invention of anesthesia in the mid­nineteenth century, surgery for cancer became more common, and was found to be sometimes successful when the tumors removed were small. For example, by 1900 the radical mastectomy was the treatment of choice for breast cancer.
X-rays were discovered by Wilhelm Roentgen in 1895, and radiation treatment of cancer began almost immediately. In the 1930s, Geoffrey Keynes pioneered the use of radiotherapy instead of or in addition to radical surgery for breast cancer. Today, radiation is used to kill many types of cancer cells such as those in bone tumors.

Chemical treatment for cancer goes at least as far back as A.D. 1000, when Avicenna gave arsenic to cancer patients. Modern chemotherapy originated in 1942, when Alfred Gilman and Louis Goodman were attempting to develop antidotes to mustard gas chemical warfare. They found that rapidly growing lymphatic tissues were especially vulnerable to mustard gases, and proposed that nitrogen mustard could be used to treat malignant lymphoma. By 1956, ten kinds of cancer were known to be treatable by drugs. Less toxic kinds of chemicals have been used to kill cancer cells, but they always have the side effect of killing ordinary cells as well ­ the usual ratio cited is six cancer cells for every one normal cell. Hence there is a limit on the strength of chemotherapy that can be used because of the large numbers of normal cells that are destroyed. Moreover, cancer cells often become resistant to different kinds of chemotherapy by mutating, so in many cases chemotherapy has only temporary benefits. The same is true for hormonal treatments often used in prostate and breast cancers.

In the 1980s, there was dramatic progress in understanding how cancers originate. According to the "two hit" theory, cancer occurs when (1) normal genes involved in cellular functions are transformed by mutation into oncogenes that excessively stimulate cell growth and (2) there is a mutation in a tumor-suppressor gene that then fails to perform its function of controlling growth. Disappointingly, the dramatic advances in understanding the molecular biology and biochemistry of cancer have been slow to yield effective treatments for the hundred or so kinds of cancers that afflict people.

Today, there are a number of novel kinds of cancer treatment in early stages of investigation. In 1960, Judah Folkman discovered that cancer tumors require a blood supply in order to grow and spread, and he began the search for drugs that would prevent this process of angiogenesis (Cooke, 2001). Numerous anti-antiogenesis drugs are now being tested in clinical trials. In a very different approach, viruses such as variants of the herpes virus have been found to have cancer-killing capabilities. Bischoff et al. (1996) discovered that an adenovirus (similar to the viruses that cause colds) can be genetically engineered to kill cells deficient in the activity of p53, which is one of the most important tumor-suppressor genes. Clinical trials are now underway to see whether these viruses can be successful in treating humans with cancer. More recently, another virus has been found to have cancer-killing potential.

3. The Reovirus

The term virus originally meant "poison," and any cause of disease could be referred to as a virus. In the 1890s, researchers found that an extract from diseased tobacco plants that had been filtered to remove bacteria could nevertheless cause disease in previously healthy plants, and it was soon found that many other diseases were associated with "filterable viruses". During the 1930s, the electron microscope was developed and it became possible to identify and describe the appearance of particular viruses. Unlike bacteria, viruses cannot reproduce on their own but are parasitic on living cells. By 1950, the field of virology had split off from bacteriology.

The first reovirus was isolated in 1951 from the feces of an Australian aboriginal child (White and Fenner, 1994). It is now recognized to be one of three serotypes of mammalian reoviruses belonging to the genus Orthoreovirus and the family Reoviridae. The term "reovirus" was coined in 1959 by Albert Sabin, known for his work on the oral polio vaccine. He formed the prefix "reo" to abbreviate "respiratory enteric orphan" because the virus was found in the body's respiratory and enteric tracts and because it was an orphan in the sense that it was not identified as the cause of any human disease. Initially, there was considerable interest in the reovirus because of its similarities with the polio virus. But ethically dubious experiments in which prisoners were injected with reovirus found that infection caused at most mild flu-like symptoms. Many people have been infected by reovirus as children with little effect more than a runny nose.

Even though the reovirus turned out to be of low medical significance, it continued to be a popular topic of research for leading virologists. Reoviruses are easy to grow and have biologically interesting properties such as consisting of double-stranded RNA. The reovirus has only ten genes, and the proteins produced by each have been isolated and characterized with respect to their effects on virulence. For example, the gene S1 produces a protein that enables reoviruses to attach to cells in the first stage of infection, but different types of reovirus attach to different kinds of cells. Reoviruses were a major focus of investigation in the Duke University laboratory of Wolfgang Joklik, where Patrick Lee arrived as a postdoctoral fellow in 1978.

4. Patrick Lee

Patrick W. K. Lee was born in China in 1945, but grew up in Hong Kong. In 1967 he moved to Edmonton, Canada, to attend the University of Alberta, receiving a B.Sc. in 1970 and a Ph.D. in biochemistry in 1978. His doctoral research concerning the biochemistry of the mengo encephalomyelitis virus was supervised by John Colter, a leading expert on that virus. Lee then received a post-doctoral fellowship from the Medical Research Council of Canada that he took to work with another leading researcher in the biochemistry of viruses, Wolfgang (Bill) Joklik, who had come to Duke University as chair of the Department of Microbiology and Immunology. Because Lee had his own funding from the Canadian government, and because Joklik was supervising numerous other postdocs and graduate students on top of his duties as department chair, Lee was given a relatively free rein to work on topics that interested him. Joklik was an intimidating supervisor, and Lee was taken aback when Joklik was not enthusiastic about Lee's proposal to do an experiment to determine which protein produced by the reovirus allows it to attach to cells. Joklik said that the experiment was not worth doing: others must have tried it already and found that it did not work. But Lee did the experiment anyway and found the reovirus cell attachment protein, a result that Lee and Joplik published in the journal Virology in 1981.

In that same year, Lee returned to Alberta as an Assistant Professor in the Department of Microbiology and Infectious Diseases at the University of Calgary. In collaboration with a series of graduate students, he continued to work on the biochemistry of reoviruses. Lee published extensively on such topics as the genetic structure of reovirus attachment gene S1 and the nature of the receptors in cells that enable reoviruses to attach to them. His research addressed fundamental questions about the biochemical mechanisms of viral development, completely independent of questions about the genesis and treatment of cancer. Lee was strongly motivated to do research equal to that of the top virologists, and became engaged in controversies with leading figures such as Bernard Fields over the nature of viral attachment. By 1991, Lee was a full professor.

5. Discovering the Reovirus-Cancer Connection

The finding that started the long path to the discovery that reoviruses can kill cancer cells came in 1992, when a graduate student, Damu Tang, proposed an experiment that Lee still describes as "silly" and "stupid." The student was interested in the role of sialic acid as a receptor on the plasma membrane of a cells. In order to attach to a cell, a virus uses attachment proteins to bind itself to such receptors. The student proposed to use an enzyme to cleave sialic acid from cells and predicted that this would block infection by reoviruses. Lee thought that the experiment would not work and would not be very interesting even if it did.

Fortunately, just as Lee had ignored Joklik's opinion of the experiment Lee had proposed as a postdoctoral fellow, Tang ignored Lee's opinion and did the experiment anyway. Because viruses reproduce rapidly ­ reovirus progeny are detectable in less than 10 hours ­ experiments like the one proposed by Tang can be done in a day or two, with very little cost in equipment. Tang found, as Lee had predicted, that the sialic acid manipulation did not produce less viral infection than occurred in the control cell medium not receiving the enzyme. The big surprise, however, was that the viral infection was reduced in the control cells! Lee was sure that the student had just made a mistake, and told him to repeat the experiment with no changes. When the student repeated the experiment with the same results, Lee told him to do it again. Yet another replication of the results convinced Lee that there was something genuinely odd going on in the control cell medium, and he began to suspect that something in it was blocking virus reproduction. He dropped everything and went to the library to try to find an answer to the question of what might be going on in the control cells.

From a journal article, Lee learned that the cells used in the experiment secrete epidermal growth factor receptor (EGFR). The cells were from the human epidermoid carcinoma cell line A431, which Lee had used in previous experiments to show that beta-adrenergic receptors, possessed in large numbers by A431 cells, are not receptors for reovirus binding. The difference between the enzyme and control conditions was not the cell lines, which were both A431, but that the control cells had been sitting around for a few days, which gave them time to secrete something that blocked viral replication. Lee initially hypothesized that reoviruses were binding to EGFR in solution in the control cell medium, which prevented them from binding to EGFR on cells that they were supposed to infect.

Lee and his students then did experiments that showed that two mouse cell lines previously known to express no EGFR were relatively resistant to reovirus infection, whereas the same cell lines became susceptible to infection with the insertion of the gene encoding EGFR. In papers published in Virology in 1993, they concluded that the process of reovirus infection is closely coupled to the cell signal pathway mediated by the EGFR (Strong, Tang, and Lee, 1993; Tang, Strong and Lee, 1993). As is common in scientific publications, this paper gave no indication of the serendipitous origins of the finding that the epidermal growth factor pathway is involved in reovirus reproduction.

At this time, Lee had not yet considered any possible connection between reovirus and cancer. He was still thinking that the virus might bind to an epidermal growth factor receptor, but further reflection about the mechanisms of reovirus binding raised doubts about whether reoviruses bind to EGFR. Lee was forced to rethink the situation, and it occurred to him that the virus might infect cells that were already prepared for infection because of an already activated chemical pathway. In a paper written in 1995, Strong and Lee lay out two alternative explanations for the augmentation of reovirus infection by functional EGFR:

The first possibility is that reovirus plays an active role by first binding to EGFR, thereby activating the tyrosine kinase activity of the latter and triggering a cell signaling cascade which is somehow required for subsequent steps of the infection process. The second possibility is that reovirus takes advantage of an already activated signal transduction pathway conferred by the presence of functional EGFR on the host cell. (Strong and Lee, 1996, p. 612).

To choose between these possibilities, Strong and Lee designed an experiment using a cell line that they heard about from a visiting speaker at the University of Calgary, H. J. Kung. Kung provided them with a cell line, NIH 3T3, which had been used extensively to assess the transforming activities of oncogenes, which are genes whose transformation can lead to cancer. Lee was still not thinking about cancer, but rather focusing on the question of whether an oncogene introduced into NIH 3T3 cells could transform them internally in a way that would show the correctness of the second possibility that reovirus infection exploits an already activated pathway.

In their experiment, Strong and Lee used the v-erbB oncogene known to be related to a growth factor receptor similar to EGFR. Decisively, it turned out that addition of this oncogene to the NIH 3T3 cells did indeed make cells that are normally resistant to reovirus infection become susceptible to infection. This result showed that the role of EGFR in the 1992 experiment was fortuitous and inconsequential: what mattered was not the attachment process of reoviruses, but the internal modification of cells that allowed the reovirus to reproduce rapidly.

The crucial question then concerned the nature of the activated signal transduction pathway within the cells that made them susceptible to infection by reovirus. Answering this question was made easier by the availability of a number of NIH 3T3-derived cell lines transformed with active oncogenes. At the end of their paper, Strong and Lee reported (1996, p. 615): "We have obtained preliminary data which suggest that activated ras alone also confers enhanced reovirus infectibility." The research motivation was still basic virology, to understand the intracellular mechanisms that promote viral replication. Ras was investigated because it was widely known to be downstream from EGFR: there is a biochemical mechanism in which EGFR affects Ras.

Patrick Lee was quoted in 1999 as saying: "I can still remember the day we found out that the reovirus could have a cancer connection. It was the most exciting day of my life" (AHMFR, 1999). The connection came from the nature of reovirus reproduction, which, as in many other viruses, leads to the destruction of the infected cells. After a reovirus gains access to a cell, it reproduces itself by using the cell's internal chemistry to replicate viral RNA which is then assembled into thousands of new viruses. These viruses then cause the cell to burst, a process called lysis, enabling the viruses to spread out to infect many other cells. The discovery that the reovirus is prone to infect cells that have an activated Ras pathway suggested immediately that the virus might be oncolytic, that is capable of killing cancer cells by infecting and bursting them. Thus by the end of 1995, Lee had the hypothesis that there might be a reovirus treatment for cancer.

Lee did not, however, immediately try to test whether reovirus could cure cancer in animals. First, he and his students set out to identify the mechanism which made cells with an activated Ras pathway susceptible to reovirus infection. The crucial protein was found to be the double-stranded RNA-activated enzyme, PKR. Inhibition of PKR activity drastically enhanced reovirus protein synthesis, and it was found that Ras-activated cells depress the activity of this protein, permitting the reproduction of reoviruses and the destruction of the cells. By 1998, Lee's team understood why reoviruses are potentially oncolytic (Strong et al., 1998). More recent research has demonstrated that a similar mechanism explains the oncolytic properties of the herpes simplex virus.

It was relatively easy to do studies that showed that reoviruses could kill many different kinds of cancer cells in test tubes, but answering the question of whether they could be effective in animals required a dramatic shift in Lee's research methodology. He launched a new line of research to determine whether injecting cancerous mice with cells infected by reovirus would have any effect on their tumors. The results were excellent, demonstrating the eradication of many tumors, even in mice with competent immune systems (Coffey et al., 1998). Another investigator got significant results in reducing tumor growth in dogs. Because reovirus is not associated with any major human diseases, it has substantial promise for treating cancers in humans, but results of the first clinical trial are not yet available. In 1998, Lee helped to start a company, Oncolytics Biotech, which has patented the use of reovirus, trademarked as Reolysin,, to treat cancer in humans. In May, 2001, the company announced that it was seeking approval for phase 2 clinical trials concerning effects of Reolysin on prostate cancer and a type of brain cancer, glioblastoma. Health Canada granted approval for the prostate cancer trial in October, 2001.

To summarize, Patrick Lee's path to the reovirus treatment of cancer was very indirect, and involved the following key steps:

1. The odd experimental result that reovirus infection was inhibited in a control cell medium.
2. The determination that epidermal growth factor receptor in cells encouraged viral reproduction.
3. The determination that EGFR was not so important, but that what mattered was an internal pathway.
4. The determination that the Ras pathway was involved.
5. The realization that the reovirus could kill cancer cells.

We can now consider what these developments can tell us about the nature of scientific discovery.

6. Serendipity and the Economy of Experimental Research

Charles Peirce, who was an active scientist as well as a philosopher, coined the phrase "economy of research" to indicate that scientific method is not just a logical matter of what to believe, but also a practical matter of what to investigate. He emphasized the need to form hypotheses that can be tested by means of economical experiments, that is ones that do not require an excess of money, time, thought, and energy (Peirce 1931-1958, vol. 5, sections 5.598-600). The economy of experiments varies dramatically from field to field. At one extreme, there is high-energy physics where an experiment can require hundreds of person-years of work and millions of dollars.

Clinical trials in medicine that evaluate the efficacy of drugs can take months or years and cost hundreds of thousands of dollars. In experimental psychology, a typical experiment requires weeks of planning, weeks of running subjects, and weeks of data analysis. Monetary costs include stipends for graduate students who run the experiments and payments to subjects. Given the high cost of conducting experiments, researchers must be very careful concerning the experiments on which they expend their resources.

In virology, the economy of research is very different. Experiments can often be done in one or two days with materials already on hand. Under such circumstances, it is not unusual for graduate students and postdoctoral fellows to ignore the advice of their supervisors concerning whether an experiment is worth doing or not. T. H. Huxley said that every now and again a scientist should perform an outrageous experiment, like blowing a trumpet at a tulip, just to see what happens. But if an experiment requires a great deal of time and expensive equipment, then outrageous experiments are out of the question.

Many discoveries in science are serendipitous, coming about even though no one was looking for them. An example is the discovery that the human stomach often contains bacteria, Helicobacter pylori, that cause ulcers. When the bacteria were discovered by an Australian physician, Robin Warren, in 1979, it was not part of a research project to find improved treatment for ulcers, but merely arose in his everyday work as a pathologist (Thagard, 1999). The finding by Warren and Barry Marshall that presence of the bacteria is highly correlated with peptic ulcers was an exciting surprise. Kevin Dunbar (2001) reports that in the scientific research groups he has studied more than half of the results of experiments were unexpected. Unexpected results do not reflect just the failure of hypotheses to survive experimental tests, but often serve to initiate new hypotheses and new experiments.

Lee's path to the reovirus treatment for cancer began with the unexpected finding that viral reproduction was inhibited in a control cell medium. Another important biological experiment where the control condition produced surprising results led to the discovery that lithium can be used to treat manic depression disorder. The researcher thought that uric acid might be involved, so he was treating guinea pigs with lithium urate. As a control, he used lithium carbonate, and was surprised to find that it calmed the guinea pigs down, showing that lithium was the key - and totally unexpected ­ factor (van Andel, 1994, p. 641; see also Roberts, 1989, and Campanario, 1996).

Judah Folkman's research on cancer angiogenesis also began serendipitously (Cooke, 2001, ch. 4). After his surgery training at Harvard, Folkman was drafted in 1960 by the U.S. Navy and assigned to work on finding a blood substitute that could be used on aircraft carriers that spend months at sea. Working with a pathologist, Fred Becker, Folkman was using rabbit thyroid glands to test the effectiveness of a hemoglobin solution as a blood substitute. They implanted cancer cells in the glands because malignant cells grow rapidly and would therefore reveal the effectiveness of the substitute. Initially, tumors implanted in the thyroids grew rapidly, but then Folkman and Becker noticed something very odd: the little tumors stopped growing as soon as they reached a size of about one millimeter in diameter. They then realized that the tumors had become dormant because they did not have any connection to the circulatory system. This was the serendipitous beginning of research on angiogenesis that has led, forty years later, to the current investigation of dozens of anti-angiogenesis drugs as potential cancer treatments. Like Patrick Lee, Judah Folkman's interest in cancer arose because of an unplanned finding in an unrelated field of biomedicine.

A field like virology, in which experiments are very economical, allows the carrying out of numerous experiments, which can in turn lead to serendipitous findings. This is what happened in the experiment that initiated the path to the reovirus treatment of cancer. Because experiments with reovirus require little time and expense, Tang was able to perform what Lee viewed as an unpromising experiment. Reoviruses do not cause serious disease, so working with them does not require the extensive safety precautions required for working with dangerous viruses such as HIV and Ebola. Thus the economy of research, in this case the ease of doing dubious experiments, contributed to the serendipitous beginning of the path to the reovirus treatment. However, as Pasteur said, chance favors the prepared mind, and I discuss below the mental mechanisms by which the surprising result of Tang's experiment was turned into a serious of hypotheses and experiments.

Another serendipitous aspect of Lee's research was the hypothesis that reoviruses bind to EGFR. This hypothesis turned out to be a red herring, but it was very useful theoretically because it helped to move Lee's theorizing away from questions of reovirus attachment towards questions about chemical processes internal to cells. It is interesting that Lee's doubts about this hypothesis arose not from experimental tests of it but from reflection on its incompatibility with what he knew about the biochemical mechanisms of reovirus attachment.
How do scientists decide what experiments to do? Peirce's idea of the economy of research suggests that such decisions might involve some kind of cost-benefit calculation, with steps such as the following:

1. Estimate the benefits of an experiment with respect to potential contribution to scientific knowledge and (perhaps) personal career advancement.
2. Estimate the costs of the experiment with respect to time, money, and energy.
3. Compare the cost/benefit ratio of the experiment with ratios for other possible experiments, and decide whether the experiment is worth doing.

However, this strikes me as an implausible model of what productive scientists actually do. For one thing, it is extremely difficult to calculate the expected benefits of an experiment. As the experiment that began the path to the reovirus treatment shows, it is often very difficult to know where an experiment will lead. Therefore, it is difficult to compare the cost/benefit ratios of different experiments. Below, in the section on emotions, I will offer a different model of experimental decisions based on emotional intuitions.

A very interesting question that has received very little attention in the history, philosophy, and psychology of science is how scientists come up with ideas for experiments. For example, how did Tang come up with the idea of using an enzyme to cleave sialic acid from cells in order to block reovirus reproduction? Sometimes, mundane experiments can originate analogically, when a scientist is familiar with the design of an existing experiment and adapts it to do a similar experiment. More generally, experimental design can be thought of as a kind of means-ends problem solving, in which there is given end, for example to test a newly generated hypothesis, from which the scientist works backward to figure out what kind of experiment might accomplish that end. Most important experiments come about, I suspect, because they are devised to answer some important theoretical question. For example, Strong and Lee devised the v-erbB experiment to find out whether reovirus reproduction takes advantage of an already activated pathway.

7. Cognitive Mechanisms: Hypothesis Formation

A central task of scientific thinking is to generate hypotheses that explain the results of observation and experiment. Peirce called the generation of scientific hypotheses abduction, which he defined as studying facts and devising a theory to explain them (Peirce 1931-1958, vol. 5, section 5.145). An abductive inference starts with a surprising and interesting fact that needs explanation, then uses background information to generate a hypothesis that provides a potential causal explanation of the interesting fact. The path to the reovirus treatment involved numerous instances of abductive inference. When Tang's experiment produced its odd result that reovirus replication was reduced in the control condition, Lee's first hypothesis was that the student had made some kind of mistake.

The structure of this abductive inference was something like:

Fact to be explained: Odd experimental result - reovirus replication reduced.
Background information: When students make mistakes in performing experiments, they often get odd results.
Hypothesis: Maybe the student made a mistake in performing the experiment.

The background information came from Lee's years of experience as an experimenter and teacher, and might have been in the general form stated here, or might have been at a lower, more particular level in the form of specific cases of students who had got odd results because of erroneous experiments.

The hypothesis that there was some mistake in Tang's experiment could easily be tested by running it again and again. Replication ruled out the mistake hypothesis, so Lee was forced to come with the alternative hypothesis that there was something unusual going on in the control cells. The odd experimental result was anomalous, but it only became a really interesting anomaly when the initial hypothesis of experimental error was ruled out. The exciting abductive inference was something like:

Fact to be explained: Odd experimental result ­ reovirus replication reduced.
Background information: Cell secretions can block viral reproduction.
Hypothesis: There is some secretion in the cell medium that is blocking viral reproduction.

Lee had no idea what the secretion might be, but found by reading journals that A431 cells secrete EGFR so it became a candidate to fill in the hypothesis. This kind of abductive inference to the existence of an unknown entity is called existential abduction (Thagard, 1988). It is common in science, for example when nineteenth-century astronomers inferred from the orbit of Uranus that there exists another planet which was later identified as Neptune.

Another major abductive mental leap came later with formation of hypotheses to explain why EGFR in host cells enhances viral reproduction. Here there were two abductive inferences:

Fact to be explained: EGFR in host cells increases reovirus reproduction.
Background information: If binding of viruses to cells is increased, then viral reproduction increases.
Hypothesis 1: Reoviruses might bind to EGFR.

However, reflection about biochemical mechanisms made this hypothesis implausible, so Lee had to generate another:

Fact to be explained: EGFR increases reovirus reproduction.
Background information: If some chemical process downstream from EGFR encourages viral reproduction, then viral reproduction increases.
Hypothesis 2: There is some chemical process downstream from EGFR encouraging viral reproduction.

As in the abductive inference that there was something in the control cell medium blocking viral reproduction, this was an existential abduction. Identification of the postulated agent came from further experiments that filled in the Ras pathway as the downstream chemical process involved in enhancing viral reproduction.

Lee's last major mental leap was more deductive than abductive. It involved the following steps:

1. Reovirus kills the cells it infects.
2. Reovirus infects cells with activated Ras pathways.
3. Cancer cells often have activated Ras pathways.
4. Therefore, reovirus may be able to kill cancer cells.

Laying the steps out like this makes the leap seemed obvious, but remember that Lee was a virologist, not a cancer researcher, so he was not particularly concerned with the oncogenic properties of Ras. The big breakthrough came when Lee and his students determined that Ras pathway activation enhanced reovirus infection, which directly suggested that reoviruses might be used to treat cancer.

8. Emotional Mechanisms

We often think of scientists as paragons of cool rationality, but they are as emotional as anyone else, and their emotionality is often a major contributor to their success (Thagard, forthcoming). Emotions are an important factor in all the phases of scientific thinking, including planning of experiments, recognition of anomalies, formation of hypotheses, and evaluation of competing theories. Aesthetic judgments that a theory or experiment is beautiful or ugly are an important class of emotional reactions (Thagard, 2000, pp. 199-203). McAllister (1996) has discussed the contribution of aesthetics to theory appraisal, and Parsons and Rueger (2000) analyze the relevance of aesthetics to experimental science.

Consider first the economy of research discussed above. I argued that the explicit cost-benefit model of choice of experiments is psychologically unrealistic. A more plausible model is the account of decision making as informed intuition (Thagard, 2001). Intuitions are emotional judgments that arise from unconscious processes that balance various cognitive and emotional constraints. For example, if I need to buy a car, I may have the feeling that I should by a particular model because it satisfies my most important goals such as being reliable and well equipped. Decision making involves both positive constraints, such as that I want the car to furnish reliable transportation, and negative constraints such as that I cannot afford to buy more than one car. These constraints includes ones that are cognitive, based on the interrelations of various compatible and incompatible beliefs that I have about cars, and also ones that are emotional, reflecting how I feel about the various factors such as style and equipment that factor into my decision. I will be intuitively satisfied in choosing a particular model of car when the choice is coherent in the sense that it maximizes satisfaction of the various cognitive and emotional constraints (see Thagard, 2000, for a theory and computational model of this kind of coherence).

Of course, intuitions may produce poor decisions, if the decisions neglect crucial cognitive and emotional constraints. For example, my decision making may be temporarily swamped by a single factor, such as seeing a car with a gorgeous color. More effective decision making requires informed intuition, in which a person has taken care to collect information about the factors relevant to the decision so that the intuitive judgment that emerges to consciousness as an emotional reaction is based on maximizing satisfaction of all the relevant constraints.

I conjecture that decisions about what experiments to do are typically made on the basis of informed intuition. When a researcher generates an idea about a possible experiment, it comes with an intuitive judgment about whether the experiment is worth doing. This intuition may be based on considerable knowledge of the relevant field and experience with similar experiments, or it may merely reflect a newcomer's need to get some research going. The intuitive judgment of the scientist who first considers the experiment may be quite different from the judgments of others, for example a supervisor or a reviewer of a grant application. Regardless of who is making the judgment, it does not come out as a cold estimate such as "This experiment has a probability of success of .7", but rather as an emotion judgment such as "This experiment is exciting" or "This experiment is pointless." Lee has been quoted as saying: "The most exciting thing about science is you wake up in the morning and say to yourself, 'Gee, what I am going to find out today?'" (AHMFR, 1999).

I described earlier how two of the important episodes in Patrick Lee's career involved an emotional mismatch between the judgments of junior and senior researchers. Experience researchers react emotionally to proposed experiments on the basis of their substantial experience with previous successful and unsuccessful experiments, as well as on the basis of their sense of the research goals of the field. The distinguished social psychologist Richard Nisbett learned about doing experiments from the reactions he got in discussions with his supervisor, Stanley Schacter. Nisbett says (personal communication, Feb. 23, 2001) "He let me know how good my idea was by grunts: non-committal (hmmm...), clearly disapproving (ahnn...) or (very rarely) approving (ah!)." Interestingly, many of Schacter's students went on to become very successful social psychologists, as have many of Nisbett's students. So when a senior researcher tells a junior one that an experiment is stupid, as Joklik did to Lee and Lee did to Tang, it is an emotional judgment based on substantial experience. Of course, such judgments are not always right, and in both those cases the youthful exuberance of the junior researchers, coupled with the relative ease of doing the experiments, led to the experiments being done anyway. Regardless of the status of the researcher, the decision whether to do or not to do an experiment is not based on any explicit cost-benefit calculation, but on an emotional reaction to the proposed plan. Depending on its simplicity and its coherence with prevailing beliefs, a researcher may make the aesthetic judgment that a proposed experiment is beautiful or ugly.

I suggested earlier that experiments are often designed to answer questions, but how are questions generated? Figure 1 is a model of the origins of scientific questions. On this model, questions arise because of curiosity about some phenomena, serendipitous findings that produce surprise, or concern with some practical need. In addition, some questions are formed because they potentially provide answers to other questions, as shown by the feedback loop connecting questioning with questioning. Patrick Lee's path to the reovirus treatment displays all these ways of generating questions. He had a long-standing curiosity about the biochemical mechanisms of viral attachment and reproduction. Serendipity was a factor when Tang's experiment produced surprising results from the control cells unexpectedly displaying reduced viral infection. Surprise is an emotion that arises from addition of information to a cognitive system that is incoherent with the representations previously in the system (Thagard, 2000, p. 194). Lee found the results of Tang's initial experiment very "odd", which I take to mean that they were surprisingly incompatible with what he already knew about reoviruses. Practical need entered into the picture later when Lee realized that reoviruses might be oncolytic and therefore relevant to the great demand for improved cancer treatments. (Lee's father had died of pancreatic cancer, which usually involves an activated Ras pathway.) Note that the inputs to the questioning process ­ curiosity, need, and surprise ­ are inherently emotional.

Figure 1. Origins of scientific questions. From Thagard (1999), p. 47.

The questioning process is crucial to the focusing of inquiry that makes useful abductions possible. Scientists do not form hypotheses about randomly chosen facts, but rather about facts that cry out for explanation through questioning driven by curiosity, need, or surprise. Figure 1 shows discovery downstream (to use the biochemists' term) from questioning, and this is certainly true of the abductive inferences described in the above section on cognitive mechanisms. Abduction is a cognitive process, but it has emotional inputs via the way that questioning helps to select the facts that are worthy of being explained. Manifestly, abduction also has emotional outputs, since the formation of a promising new hypothesis can be an intensely exciting event in the life of a scientists. When I interviewed Patrick Lee, he used words like "amazing," "excited," "astounding," and "this is it" to describe his reaction to the new ideas that emerged in the course of his research between 1991 and 1996.

Figure 2 shows a more general model of the role of emotion in scientific thinking. As the story of Patrick Lee and the reovirus illustrates, the mental mechanisms by which scientific discoveries are made include emotional as well as cognitive processes. Lee's interest and curiosity led him to generate questions such as what was going on in the control medium in Tang's surprising experiment. The attempt to generate answers sometimes led to surprises such as the implausibility of the initial hypothesis about reoviruses binding to EGFR. The evaluation of answers sometimes led to happiness, as when Lee determined that the Ras pathway crucially influenced reovirus reproduction.

Figure 2. Model of the role of emotions in scientific cognition. From Thagard (forthcoming).

9. Conclusion

In sum, Patrick Lee's path to the reovirus treatment provides a fine illustration of the complex process of scientific development. It was not a simple matter of having an idea and testing it, but required a series of developments that combined experiments (both failed and successful), information acquisition, and hypothesis generation. The mental mechanisms involved in this study were diverse, ranging from decision processes for deciding what experiments are worth doing to abductive processes by which hypotheses are generated to explain surprising facts. Emotional processes were not just a byproduct of the cognitive processes, but were an essential contributor to those processes. The mental mechanisms of scientific thinkers involve interconnected cognitive/emotional processes that support anomaly recognition, experiment planning, and hypothesis formation.

At this writing (October, 2001), it is impossible to say whether the research of Lee and his students will constitute a breakthrough in treating human cancers. It may turn out that the difficulty of delivering reoviruses to cancerous tumors and the robustness of the human immune system will render the reovirus treatment of cancer ineffective. But it is also possible that clinical trials over the next year or two will show that the reovirus treatment provides a useful weapon in the war against cancer. In either case, Patrick Lee's discoveries about the mechanisms of reovirus infection stand as valuable additions to scientific knowledge. This paper has combined historical, philosophical, and psychological perspectives to describe how those additions came about.

Acknowledgements: I am extremely grateful to Patrick Lee for very informative telephone conversations, as well as for comments on a previous draft. Thanks to James McAllister for helpful suggestions, and to the Natural Sciences and Engineering Research Council of Canada for research support.


AHMFR (1999) Reovirus: Cancer fighter ­ 1999. Alberta Heritage Foundation for Medical Research Triennial Report,, consulted 1 October 2001.
BISCHOFF, J. R. , KIRN, D. H., WILLIAMS, A., HEISE, C., HORN, S., MUNA, M., NG, L., NYE, J. A., SAMPSON-JOHANNES, A., FATTAEY, A., & MCCORMICK, F. (1996). An adenovirus mutant that replicates selectively in p53-deficient human tumor cells, Science, 274, pp. 373­376.
CAMPANARIO, J. M. (1996) Using Citation Classics to study the incidence of serendipity in scientific context, Scientometrics, 37, pp. 3­24.
COFFEY, M. C., STRONG, J. E., FORSYTH, P. A., & LEE, P. W. K. (1998). Reovirus therapy of tumors with activated Ras pathway, Science, 282, pp. 1332­1334.
COOKE, R. (2001) Dr. Folkman's War: Angiogenesis and the Struggle to Defeat Cancer. (New York, Random House).
DUNBAR, K. (2001) What scientific thinking reveals about the nature of cognition, in K. CROWLEY & C. D. SCHUNN & T. OKADA (Eds.), Designing for Science: Implications from Everyday, Classroom, and Professional Settings (Mawah, NJ: Erlbaum) pp. 115­140.
MCALLISTER, J. W. (1996) Beauty and Revolution in Science (Ithaca: Cornell University Press).
OLSON, J. S. (1989) The History of Cancer: An Annotated Bibliography (New York, Greenwood Press).
PARSONS, G. G., & RUEGER, A. (2000) The epistemic significance of appreciating experiments aesthetically, British Journal of Aesthetics, 40, pp. 407­423.
PEIRCE, C. S. (1931-1958) Collected Papers (Cambridge, MA: Harvard University Press).
ROBERTS, R. M. (1989) Serendipity: Accidental Discoveries in Science (New York, Wiley).
STRONG, J. E., COFFEY, M. C., TANG, D., SABININ, P., & LEE, P. W. K. (1998) The molecular basis of viral oncolysis: Usurpation of the Ras sigaling pathway by reovirus, The EMBO Journal, 17, pp. 3351­3362.
STRONG, J. E., & LEE, P. W. K. (1996) The v-erbB oncogene confers enhanced cellular susceptibility to reovirus infection, Journal of Virology, 70, pp. 612­616.
STRONG, J. E., TANG, D., & LEE, P. W. K. (1993) Evidence that the epidermal growth factor receptor on host cells confers reovirus infection efficiency. Virology, 197, pp. 405­411.
TANG, D., STRONG, J. E., & LEE, P. W. K. (1993). Recognition of the epidermal growth factor receptor in reovirus. Virology, 197, pp. 412­414.
THAGARD, P. (1988) Computational Philosophy of Science (Cambridge, MA, MIT Press).
THAGARD, P. (1999) How Scientists Explain Disease (Princeton, Princeton University Press).
THAGARD, P. (2000) Coherence in Thought and Action (Cambridge, MA: MIT Press).
THAGARD, P. (2001) How to make decisions: Coherence, emotion, and practical inference, in E. MILLGRAM (Ed.), Varieties of Practical Inference (Cambridge, MA: MIT Press) pp. 355­371.
THAGARD, P. (forthcoming) The passionate scientist: Emotion in scientific cognition, in P. CARRUTHERS, S. STICH & M. SIEGAL (Eds.), The Cognitive Basis of Science (Cambridge: Cambridge University Press).
VAN ANDEL, P. (1994) Anatomy of the unsought finding, serendipity: Origin, history, domains, traditions, appearances, patterns, and programmability, British Journal for the Philosophy of Science, 45, pp. 631­648.
WHITE, D. O., & FENNER, F. J. (1994) Medical Virology (San Diego: Academic Press).

Note on contributor: Paul Thagard is Professor of Philosophy and Director of the Cognitive Science Program at the University of Waterloo. His most recent books are Coherence in Thought and Action (MIT Press, 2000) and How Scientists Explain Disease (Princeton University Press, 1999).
Correspondence: Paul Thagard, Philosophy Department, University of Waterloo, Waterloo, Ontario, Canada, N2L 3G1.


Back to science and medicine articles.

Back to emotion articles table of contents.

Back to Computational Epistemology Laboratory.


This page created October 12, 2001.