The “Scientific Method”

Given that a Master’s Thesis is required to demonstrate the capacity for sustained rational argument, it is natural to look for examples of rational arguments. A model that is frequently cited is the “scientific method”, and so in this lecture we will attempt to outline its characteristics: it is in fact less a method than a style of discourse that by tradition is used in scientific and engineering disciplines. However, so successful has Science been in the last century or two that strong passions are aroused on every side of the argument as soon as any attempt is made to write about the “scientific method”. Some history of Science is inescapable to attempt to explain the current status of the scientific method, and to introduce some of the terms that are commonly used. We end with a brief series of prescriptions on how to present an argument in a scientific style, according to current (anno 2000) conventions.

Induction and Deduction

Although the origins of many chemical processes can be traced back to ancient civilisations, and engineering mathematics to Archimedes, Francis Bacon was one of the first to set out a scientific approach:

The course I propose for the discovery of sciences is such as leaves little to the acuteness or strength of wits, but places all wits and understandings nearly on a level. [Bacon, F. Novum Organum, 1620]

That is, the path traced by scientific thought is one that anyone can retrace: it is a sort of public knowledge. Science has many techniques which need to be mastered by those who would use them, but only one method, which, strangely, scientists do not go out of their way to learn. Although discovery is rarely a rational or orderly process, what has been learned needs to be written up in a methodical way, so that science is advanced by rational argument and the orderly presentation of evidence.

But what is a rational argument? Following the obvious errors of the Athenian school of philosophy (Socrates, Plato) Aristotle studied the nature of argument:

We must distinguish how many species there are of dialectical arguments. There is, on the one hand, induction, on the other, reasoning [Aristotle, Topics, c. 330 BC].

First, deductive argument is represented by reasoning (i.e. syllogism): “All men are mortal. Socrates is a man. Therefore, Socrates is mortal.” Here it is logically impossible to affirm the premises while denying the conclusion, and if there are generally accepted premises that can apply to the current topic of discussion, then conclusions can be deduced. However, the only way to arrive at the general principles that might serve as premises, appears to be induction, which argues from individuals to universals: “Gold, copper, lead etc, are ductile. Gold, copper, lead etc, are metals. Therefore, all metals are ductile.” This type of argument often leads to incorrect results (mercury is not ductile), and many famous scientists have fallen into this sort of error. For example, Lavoisier (1789) claimed that all acids contained oxygen, although he admitted that he had been unable to decompose “the acid of sea salt”, proved in 1810 by Humphrey Davy to be hydrochloric acid.

Bacon (1605) diagnosed the problem as relying on merely an enumeration of instances, and proposed instead eliminative induction, “separating nature by proper rejections and exclusions”, eliminating all of alternatives until the correct one is found. This method works well with detective stories, but in real science it has produced many failures, such as Lord Kelvin’s attempt to measure the age of the Earth by comparing its temperature with all available sources of energy. Since he did not know about radioactivity, he concluded that the Earth could not be more than ten million years old, despite the claims of geologists that this could not account for the fossil record.

Evidently, other forms of induction were desirable. John Stuart Mill (System of Logic, 1843) drew on earlier methods by medieval schoolmen of resolution and composition by which complex phenomena could be first resolved into their composite elements so that appropriate causes could be identified. He devised four experimental methods: of agreement (seeking a common causal factor in disparate observations), of difference (where possible causes are successively eliminated), of concomitant variation (where possible causative factors are increased and decreased to observe their effect), and of residues (where results and their known causes are removed from theobservational account). But the blind application of such methods would lead to inappropriate results. Suppose that a series of fires occurred at a chemical works: investigation using these rules would probably conclude that the common cause was the presence of oxygen: but this information would be of little use in preventing a recurrence.

Philosophers continued to develop more and more complex methodologies for induction, but these became more and more remote from the everyday practice of scientists, and to the extent that such philosophers claimed to present a universal scientific method, just as subject to the easy counterexample as their predecessors. For example, Hume (Treatise of Human Nature) pointed out that induction also required an assumption that it was valid to argue “All observed A’s are B’s. Therefore, All A’s, future, past and distant, are B’s.” The next step was to reintroduce deduction, to allow hypotheses to be tested and rejected.

The Hypothetico-Deductive Method

A more promising approach is to attempt a likely-looking explanation of events, with Plato (Timaeus, 29):

Do not therefore be surprised, Socrates, if on many matters concerning the gods and the whole world of change we are unable in every respect and on every occasion to render consistent and accurate account. You must be satisfied if our account is as likely as any, remembering that both I and you who are sitting in judgement on it are merely human, and should not look for anything more than a likely story in such matters.

So, for example in astronomy, we should aim to construct a story within which the current state of affairs would be a natural outcome. René Descartes (Principia Philosophiae, 1644) coined the word hypothesis for likely stories that appeared to account for many observations. The danger is then that people are likely to have several likely stories at hand, each capable of explaining some observations, but contradicting each other, such as the wave and particle theories of elementary particles in physics before 1923.  Thomas Reid (c. 1785) argued that it was easy for Descartes and Newton to be misled by analogy and love of simplicity, and throughout the history of science many mathematical formulae have been used to predict behaviour of poorly-understood phenomena for no reason beyond these.

If however, the hypothesis explained all the observed facts and then moreover were to be used successfully to predict something never hitherto suspected, it should command respect. Whewell (1840), who is credited with coining the word “Scientist”, proposed this test in his version of inductivism. Eddington in 1919 thus claimed a spectacular proof of Einstein’s Special Theory of Relativity (1915) when the precession of the orbit of Mercury was in accordance with the theory. However, as Gjertsen observes, if the precession of the orbit had been measured in 1914 (as it nearly was), Einstein’s theory could not have used the data as such a proof: it would merely have been another observation that the hypothesis would have to account for.

A system of hypotheses that provides a satisfying set of causal explanations is often called a theory. In the early twentieth century there was great confidence that the scientific method as so far described led to the discovery of truth about the real world. Popper’s work on scientific theories was hugely popular, but he and his followers (such as Lakatos) made a serious error in his use of the word falsification, and his confusion of the methods described above with the mathematical notion of proof by contradiction. For Popper, a successful theory was to be regarded as true, until proved false by an experiment whose results did not agree with the predictions of the theory. Thus Einstein “disproved” Newtonian physics. Few people today consider this approach to be useful: Newtonian physics continues to provide useful results for velocities that are much less than that of light, so that theories are not rejected or disproved, merely qualified by reference to more successful theories.

The hypothetico-deductive method, despite its critics, continues in use today. It does not seem to matter now where the hypothesis comes from, but it should normally have some family resemblance to previous successful hypotheses in the area of study. All existing observations in the area it is attempting to explain should be deducible from the hypothesis. Its claim to acceptance should then be subject to its providing predictions for further observations; and when these can no longer be deduced from the hypothesis, the scientific community should start to look for better hypotheses.

It has been widely noted that the notion of “all existing observations” is an elastic one. It frequently occurs that previous observations are discarded along with the theory, and for Kuhn this particularly happens during “scientific revolutions” that cast doubt on the methods, concepts and theories of the past.

A comprehensive theory, after all, is supposed to contain also an ontology that determines what exists and thus delimits the domain of possible facts and possible questions. New views soon strike out in new directions and frown upon the older problems (what is the base upon which the earth rests? What is the specific weight of phlogiston? What is the absolute velocity of the earth?) and the older facts ([evidence of witchcraft, demonic possession, the properties of phlogiston and the ether, etc]) which so much exercised the minds of earlier thinkers. [Feyerabend, p. 161]

As in the previous section, it is demonstrable that scientific practice does not follow the rules of scientific method. Critics such as Kuhn and Feyerabend go further and make an impressive case that the rules of “critical rationalism” should not be strictly observed in scientific practice.

To start with we have seen, though rather briefly, that the actual development of institutions, ideas, practices, and so on, often does not start from a problem but rather from some extraneous activity, such as playing, which, as a side effect, leads to developments which later on can be interpreted as solutions to unrealized problems… Secondly, .. a strict principle of falsification, or a ‘naïve falsification’ as Lakatos calls it, would wipe out science as we know it and would never allow it to start… Finally, we have by now seen quite distinctly the need for ad hoc hypotheses .. these specify possible explananda and explanatia, and thus determine the direction of future research. [Feyerabend, p.160-3]

In other words, the “scientific method” is a good servant but a poor master: it provides a good paradigm for writing up scientific work but is not a reliable investigative tool.

Modern vs Postmodern, Subjective vs Objective

Modern science and philosophy took as their founding principle the idea that the laws of physics explained everything in the physical world: and they continued to assert this even in the face of the huge changes that occurred in physics and mathematics in the early twentieth century (relativity destroying the concept of simultaneity, quantum mechanics with its principle of indeterminacy, the discovery in mathematics of the existence of undecidable propositions). Late modernism made the further concession that maybe there were laws of chemistry and biology that were not reducible to laws of physics. Postmodernism notes these failures, and is generally suspicious of grand narratives, universal methods, absolute realities or objective truths. The best we can do is to present our ideas in as rational a manner as possible.

If “scientific method” means merely being rational in some given area of inquiry, then it has a perfectly reasonable “Kuhnian” sense – it means obeying the normal conventions of your discipline, not fudging the data too much, not letting your hopes and fears influence your conclusions unless those hopes and fears are shared by all those who are in the same line of work, being open to refutation by [experiment], not blocking the road of inquiry. [Rorty, p.195]

Indeed in his writings Rorty replaces the pretence that the goal of science is objective truth with more social notions of solidarity hinted at by the above quotation. He and others have argued persuasively that words like “objective” are merely compliments that we pay to statements and lines of argument that we find convincing, rather then methods of reaching absolute truth.

On the other hand, rational discourse does attempt at least to be independent of the personal prejudices of its protagonists, and seeks to persuade anyone who suspends their own personal opinions for long enough to follow the same road, that the arguments offered could lead to valid conclusions; that the experiments performed are convincing and in principle repeatable, and do not depend on special or extraneous circumstances such as the identity of the observer. If this is what “objective” is taken to mean, then critics such as Rorty and Feyerabend would have little complaint.

There is an alternative view of the objective/subjective issue:

While objective thought translates everything into results, and helps all mankind to cheat, by copying these off and reciting them by rote, subjective thought puts everything in process and omits the results [Kierkegaard, 1840, p.68]

In other words, writing becomes objective merely by being set out in order: while it may have the appearance of universal truth we all know such claims are spurious. Writing is subjective if it merely describes process and does not tell us the results.

Some guidelines for your project

The hypothetico-deductive method is so well established as an acceptable research approach that it requires no justification if you include it in your thesis. Your examiners will be delighted if you present your thesis in a form where a hypothesis is stated early on and subjected to analysis and experiment. Such a style is undoubtedly easiest in scientific or quantitative social research. It is less easy to work with in qualitative research, but here the hypothesis may usefully be replaced with questions that the research should seek to answer.

The hypothesis (or the questions) should recall or make use of the materials you refer to in your review of the literature: it should at least seem that the materials you review in the early chapters of your thesis are what led to your hypothesis or form the background to your inquiry.

The hypothesis and your answers to the questions should take account of the results of previous work, and in your suggestions for future work you should draw attention to some predictions that your research might lead towards.

If your project is of the software development form, there are often hypotheses that can be used: “this technique will be appropriate”, “software designed in this way will meet the requirement in a cost-effective way” etc. which the development of the software and subsequent testing will largely answer.

Beyond that, there are many lessons to be learnt from the origins and use of the scientific method as outlined and commented on in the above sections and the references. As you examine each paragraph you can review what you have written using questions such as “have I covered all the angles?”, “why these examples?”, “why these interviewees?”, “are all the suspects included?”, “is this induction or deduction?”, and “how could my hypothesis be disproved?”

References

Barry Barnes, “Thomas Kuhn”, in Quentin Skinner (ed.), The Return of Grand Theory in the Human Sciences, Cambridge, 1985

Paul Feyerabend, Against Method, New Left Books 1975/Verso 1988

Derek Gjertsen, Science and Philosophy: Past and Present, Penguin Books, 1989

Douglas R Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid, Penguin Books, 1980

Søren Kierkegaard, S: Concluding Unscientific Postscript to the Philosophical Fragments: A Mimic-Pathetic-Dialectic Composition: An Existential Contribution by Johannes Climacus, Swenson, D F, Lowrie, W, eds, Princteon, 1941 (orig publ Copenhagen1840)

Thomas Kuhn, The Structure of Scientific Revolutions, University of Chicago, 1962

Antoine Lavoisier, Elements of Chemistry, (orig publ as Traité élémentaire de Chimie, 1789) Dover, New York, 1965.

Karl Popper, The Logic of Scientific Discovery, (orig publ as Logik der Forschung, Vienna 1934) Hutchinson, London, 1959

Richard Rorty, Consequences of Pragmatism, Harvester Wheatsheaf, 1991

Stanford Encyclopedia of Philosophy entry on Whewell: http://plato.stanford.edu/entries/whewell/

William Whewell, Philosophy of the Inductive Sciences, Thoemmes Press, Bristol, 1999 (orig publ 1840).