Peter Medawar was by all accounts a brilliant and witty man. He won the Nobel Prize in 1960 for his work on graft rejection, which had a huge impact on the field of immunology. His wit and personality come through in this little volume, originally given as a series of lectures in 1968. I found this book to be rich in ideas, and still relevant today. I begin with a summary of the book (with comments), then describe its place in my research. In many cases I will quote him directly, mainly because I enjoy his prose.
Medawar states in the preface that the lectures “began in my mind in the form of a question: why are most scientists completely indifferent to – even contemptuous of – scientific methodology?”(p. vii)*. It wasn’t until this past year that I learned the definition of methodology. It is not, as I had previously thought, a fancy word for methods, or a set of methods. It is the analysis of methods. Of course individual methods are analysed carefully for effectiveness (eg PCR or flow cytometry), but that’s not what Medawar is talking about. He’s referring to “The Scientific Method” and the way we think when we do science.
Chapter 1: The Problem Stated
In the first chapter, Medawar explains the question. What exactly do scientists (specifically biologists) actually do to make scientific discoveries? He argues that most scientists are themselves unable to answer this question. The few who have tried either produce misrepresentations or are not scientists at all but lawyers, historians or sociologists (the notable exception being William Whewell, a biologist, who Medawar refers to repeatedly). Nevertheless, scientific discovery continues! So why bother with scientific methodology at all? He suggests it would address questions of (1) validation, (2) reducibility and emergence, and (3) causality, which are of interest to all sciences (even the social ones).
Chapter 2: Mainly About Induction
In this chapter Medawar argues that induction, long referred to as the core of the scientific method, simply isn’t. This is not a new or unique argument, and he explains why: “Induction, then, is a scheme or formulary of reasoning which somehow empowers us to pass from statements expressing particular ‘facts’ to general statements which comprehend them. These general statements (or laws or principles) must do more than merely summarize the information contained in the simple and particular statements out of which they were compounded: they must add something … Inductive reasoning is ampliative in nature. … This is all very well, but the point to be made clear is that induction, so conceived, cannot be a logically rigorous process. … No process of reasoning whatsoever can, with logical certainty, enlarge the empirical content of the statements out of which is issues.” (pp. 23-4)
One of Medawar’s problems with induction is that it doesn’t account for the critical use of experiments. He describes four different types of experiments:
- Inductive or Baconian, what I would call exploratory experiments, of the type ‘I wonder what would happen if…’. These are not critical experiments. They are meant to “… nourish the senses, to enrich the repertoire of factual information out of which inductions are to be compounded.” (p. 35)
- “Deductive or Kantian experiments, in which we examine the consequences of varying the axioms or presuppositions of a scheme of deductive reasoning (‘let’s see what would happen if we take a different view’).” (p. 35)
- “Critical or Galilean experiments: actions carried out to test a hypothesis or preconceived opinion by examining the logical consequences of holding it.” (p. 37)
- “Demonstrative or Aristotelian experiments, intended to illustrate a preconceived truth and convince people of its validity.” (p. 37)
His argument is that multiple stages of experimentation are necessary in the course of original research, and critical experiments (type 3) are necessary to progress beyond “academic play” (p. 38). The version of the scientific method I was taught in school described only critical experiments – I consider that to be just as serious a misrepresentation of science methodology as only including Inductive or Baconian experiments.
Looking at this list from a modern perspective, I wonder whether we are able to distinguish so clearly between critical or demonstrative experiments. What exactly separates the two? Is it intent? If so, how are we to judge another’s experiments simply by reading a paper in a journal? The conclusion I have drawn after reflecting on this list is that published papers may be a good way of disseminating results, but they are very poor at representing the methodology of science. As Medawar says later in the book, “The critical process in scientific reasoning is not … wholly logical in character, though it can be made to appear so when we look back upon a completed episode of thought.” (p. 53 ). In other words, papers present a logical progression that is seen only in hindsight, and doesn’t reflect the reality of research. Baconian experiments are necessary to embark on any new area of research, but they are rarely published unless something extraordinary is stumbled on.
Medawar goes on to explain the specific shortcomings of induction as a methodology, at the same time highlighting the requirements of a good methodology. This is his summary at the end of the chapter:
- “Inductivism confuses, and a sound methodology must distinguish the process of discovery and of justification.
- The evidence of the senses does not enjoy a necessary or primitive authenticity. The idea, central to inductive theory, that scientific knowledge grows out of simple unbiased statements reporting the evidence of the senses is one that cannot be sustained.
- A sound methodology must provide an adequate theory of special incentive – a reason for making one observation rather than another, a restrictive clause that limits observations to something smaller than the universe of observables.
- Too much can be made of matters of validation. Scientific research is not a clamor of affirmation and denial. Theories and hypotheses are modified more often than they are discredited. A realistic methodology must be one that allows for repair as readily as refutation.
- A good methodology must, unlike inductivism, provide an adequate theory of origin and prevalence of error…
- … and it must also make room for luck.
- Due weight must be given to experimentation as a critical procedure rather than as a device for generating information; to experimentation as a method of discriminating between possibilities.”(pp. 40-41)
Chapter 3: Mainly About Intuition
In the final chapter, Medawar makes a case for a “hypothetico-deductive” scheme of science methodology, originating from many thinkers including Kant, Robert Hooke, Stephen Hales and Robert Boscovich, and advocated in Medawar’s time by Karl Popper. He details how each of the 7 requirements for a better methodology are met by this model, but I won’t get into that here. His focus, and I think the more interesting aspect of the chapter, is on the role of intuition or creativity in this model. “Scientific reasoning is an exploratory dialogue that can always be resolved into two voices or two episodes of thought, imaginative and critical, which alternate and interact.” (p. 46).
Again, he describes four types of creativity (not ruling out the existence of more): deductive, inductive, wit, and experimental flair. The details are less important than the conclusions he draws from their existence: “… an imaginative or inspirational process enters into all scientific reasoning at every level.” (p. 55). “That ‘creativity’ is beyond analysis is a romantic illusion we must now outgrow. It cannot be learned perhaps, but it can certainly be encouraged and abetted. We can put ourselves in the way of having ideas, by reading and discussion and by acquiring the habit of reflection, guided by the familiar principle that we are not likely to find answers to questions not yet formulated in the mind.” (p. 57).
I loved reading this book. Every section seemed to cut right to the heart of an issue and reveal it starkly, I think in large part to Medawar’s lovely style of writing. The questions he identifies as important across scientific disciplines (validity, reduction and emergence, and causality) remain relevant today, and I have a lot of personal interest in them. Beyond that, the overall premise of the book is closely related to my research. A large part of my work at the moment is focused on second order science. As defined on the website: “First-Order Science is the science of exploring the world. Second-Order Science is the science of reflecting on these explorations.” It seems to me that Medawar is doing exactly that when he talks about science methodology.
Throughout the text Medawar advocates self-reflection, or reflexivity, in science. Unfortunately, most scientists remain as unconcerned with such things today as they were 50 years ago – that’s the domain of philosophers of science – despite the potential implications. What would be the benefit of engaging with science methodology in the way Medawar did? He says: “Most scientists receive no tuition in scientific method, but those who have been instructed perform no better as scientists than those who have not.” (p. 8). Could we change that state of affairs? Could we teach ourselves to be better scientists if only we could describe what we are doing?
In the past year I have done a lot of reading about systems theory, with a bit of complexity science and cybernetics thrown in there. A key part of General Systems Theory, as originally defined by Ludwig von Bertalanffy, is the idea that there are patterns and general rules that can be found and applied in systems across all disciplines. In that sense, it is a transdisciplinary theory. Cybernetics, as defined by Norbert Wiener, is the study of communication and control, now referred to as the study of regulatory systems. It is also transdisciplinary, and closely related to systems thinking in intellectual lineage. Given that background context, I was very excited to read the following passage (pp. 54-55):
“There is nothing distinctively scientific about the hypothetico-deductive process. It is not even distinctively intellectual. It is merely a scientific context for a much more general stratagem that underlies almost all regulative processes or processes of continuous control, namely feedback, the control of performance by the consequences of the act performed. … scientific behavior can be classified as appropriately under cybernetics as under logic.”
This observation, combined with his repeated suggestion of reflexivity, shows that Medawar was thinking in terms of second-order science. This is a lovely example of synchronicity, or the same idea occurring separately in many places at once, because a group of cyberneticians described second-order cybernetics shortly after.
The more I think about what Medawar wrote, the more I link it to the rest of my research. I will certainly be referring to Induction and Intuition in Scientific Thought in future posts, and I highly recommend it if you have any interest in the way scientists think.
*All page number citations refer to:
Medawar, P. B. 1969. Induction and Intuition in Scientific Thought. Vol. 75. Memoirs of the American Philosophical Society. Trowbridge & London: Redwood Press Limited.