The methodology of scientific research programmes
The methodology of scientific research programmes, Imre Lakatos, Philosophical papers, volume I, edited by John Worrall and Gregory Currie, Cambridge University Press, 1995 printing, ISBN 0-521-28031-1, pbk.
Imre Lakatos was one of the major modern philosophers of science. His name and work is often placed in contrast with that of Popper, Kuhn, and Feyerabend. Volume I of the collected papers deals with philosophy of science; volume II contains collected papers on the philosophy of mathematics.
Volume I consists of an introduction and five papers. The introduction, Science and Pseudoscience, was written in 1973 and was delivered as a radio address. The papers are:
1. Falsification and the methodology of scientific research programmes
The entirety of this collection is worth reading and rereading very carefully. The first paper is THE paper, his major position paper. In this review I will go through it and the introduction in some detail.
The introduction raises the question: How do we tell science from pseudoscience? It begins with a survey of proposed answers and the problems with those proposals. Thus, some philosophers have drawn the line by saying “a statement constitutes knowledge if sufficiently many people believe it sufficiently strongly.” It should be clear that this will not do for delimiting scientific knowledge. He then goes on to quote Hume:
If we take in our hand any volume; of divinity, or school metaphysics, for instance; let us ask, does it contain any abstract reasoning concerning quantity or number? No. Does it contain any experimental reasoning concerning matter of fact and existence? No. Commit it then to the flames. For it can contain nothing but sophistry and illusion. [p2]This sounds very well until one asks, what is the nature of this “experimental reasoning”. Lakatos continues:
But what is ‘experimental reasoning’? If we look at the vast seventeenth-century literature on witchcraft, it is full of reports of careful observations and sworn evidence – even of experiments. Glanvill, the house philosopher of the early Royal Society, regarded witchcraft as the paradigm of experimental reasoning. We have to define experimental reasoning before we start Humean book burning.An early answer, favored by Newton who believed himself to have done such, is that one proves theories by deducing them from facts. This is readily seen to be impossible; one cannot, in general, deduce general laws from a finite number of facts. Lakatos attributes this false belief in provability to the inheritance of attitudes taken over from theology where provability is in the cards because theology starts with a presupposition of certain knowledge.
The presumption that provability was attainable was buttressed by the enormous success of Newtonian mechanics – scientists believed that Newton had deciphered God’s ultimate laws. Then came Einstein and provability was recognized as a mirage, raising (reraising) the question:
If all scientific theories are equally unprovable, what distinguishes scientific knowledge from ignorance, science from pseudoscience? [p3]One twentieth century answer was “inductive logic” wherein theories were to be rated by their mathematical probability of satisfying the available total evidence. This sounds plausible and has the attractive feature of providing a measurement of quality. Popper, however, argued that the mathematical probability of any theory whatsoever, regardless of the amount of evidence, is zero. Popper in turn proposed the falsification criterion:
A theory is ‘scientific’ if one is prepared to specify in advance a crucial experiment (or observation) which can falsify it, and it is pseudoscientific if one refuses to specify such a ‘potential falsifer’. [p3] In other words, we cannot prove scientific theories but we can disprove them. A pseudoscientific theory is one which admits of no means of disproof.
A major difficulty with Popper’s criterion is that science doesn’t work that way. Scientific theories have tenacity; theories are not jettisoned immediately because of facts that which contradict them. Sometimes rescue hypotheses are constructed; sometimes anomalies are simply ignored or are set aside to be considered later. Quite often crucial falsifying experiments are only seen to be such well after the fact.
If Popper has been falsified, what then? Lakatos says that Kuhn in turn suggests that “a scientific revolution is just an irrational change in commitment, that it is a religious conversion”. This, in my opinion, is not at all a proper interpretation of Kuhn’s theses although Kuhn is often read that way by post-modernists.
Lakatos proposes the a theory of research programmes. He remarks that research programmes are the unit of scientific achievement rather than isolated hypotheses. A research programme typically a hard core, a protective belt of auxiliary hypotheses, and a heuristic, i.e., problem solving machinery. Thus, in the Newtonian programme, the laws of motion and the universal law of gravitation are the hard core. Anomalies in the motion of planets are dealt with by considering factors that may affect the apparent motion, e.g., refraction of light or the existence of a hitherto unknown planet. The problem solving machinery is the vast body of classical mathematical physics.
Research programmes, scientific or pseudoscientific, have, at any stage, both undigested anomalies and unsolved problems. “All theories, in this sense, are born refuted and die refuted.” Lakatos then makes a dubious move, to wit:
But how can one distinguish a scientific or progressive programme from a pseudoscientific or degenerating one? [p5] The move, here, is to identify “scientific and progressive” and “pseudoscientific and degenerating”. He goes on to nominate “predicting new facts” as a major criterion for distinguishing between progressive and degenerating research programmes:
Thus, in a progressive research programme, theory leads to the discovery of hitherto unknown novel facts. In degenerating programmes, however, theories are fabricated only in order to accomodate known facts. Has, for instance, Marxism ever predicted a stunning novel fact successively? Never. [p5]He was doing so well, up to his rhetorical query about Marxism. He followed this with a list of failed predictions of Marxist theory. In point of fact, though, Marx made a number of novel predictions which panned out. Thus, Marxist theory predicted (and it was far from obvious at the time) the consolidation and merging of large firms and the increasingly convulsive cycle of booms and depressions – capitalism followed the predictions of Marxist theory up to the great depression. This is not to say that Marxist theory was correct (he was, after all, an economist) or that the Marxist research programme did not degenerate; in fact, it did because it fundamentally could not take into account the reaction of the industrial nations to economic crises.
Nor is it accurate to say that pseudosciences do not predict novel facts, even stunning natural facts. Velikovsky, for example, predicted that Venus would be hot and that Jupiter would be a radio source. For that matter, psychic hotline psychics have been known to make stunningly accurate predictions from time to time.
Lakatos, like many philosophers of science, tends to focus on physics, a focus that tends to be misleading. One of the striking features of physics as a science is the reduction of the domain of phenomena to be considered. This reduction is an enabler of the possibility of precise prediction.
Falsification and the methodology of scientific research programmes
This paper is the one that establishes the main force of Lakatos’s argument. It begins with the observation that prior to the twentieth century and Einstein “knowledge meant proven knowledge – proven either by the power of the intellect or by the evidence of the senses.” in response to Einstein’s results the notion that scientific knowledge is proven knowledge has pretty much been abandoned.
He then spends a few paragraphs contrasting Popper and Kuhn before launching into a taxonomy of philosophical positions.
“According to the ‘justificationists’ scientific knowledge consisted of proven propositions.” Lakatos distinguishes between Classical Intellectualists who admitted powerful sorts of extralogical proofs, e.g., by revelation and intuition, and Classical Empiricists who admitted as axioms only a hard core of empirical “proven facts”. The latter necessarily augmented classical deductive logic with “inductive logic”. In the long run justificationism failed.
This approach treats scientific knowledge as “highly probable but not provable”. As noted above Popper established (according to Lakatos) that this does not work. It would have been nice if Lakatos had fleshed this assertion out – in ordinary parlance one speaks of various hypotheses being more or less probable.
In turn Popper introduced dogmatic falsification:
Scientific honesty then consists of specifying, in advance, an experiment such that if the result contradicts the theory, the theory have to be given up. [p13]Lakatos argues at some length that dogmatic falsification is untenable. The essence of the matter seems to be that the line between “experimental fact” and “theory” is not absolute. Lakatos also argues that knowing what is “fact” and what is “theory” presupposes that there is a natural psychological (perceptual) border between them. In turn, this means that one has to be in one’s right mind to make the distinction. In his argument I particularly liked:
… Indeed, all brands of justificationist theories of knowledge which acknowledge the sense as a source (whether as one source or as the source) of knowledge are bound to contain a psychology of observation…. All schools of modern justificationism can be characterized by the particular psychotherapy by which they propose to prepare the mind to receive the grace of proven truth in the course of a mystical experience. [p15]Irrespective of the difficulties of grounding knowledge in the senses there is a methodological problem which is fatal (and is central in Lakatos’s treatment): Predictions and theory are always subject to an “all other things being equal” clause. Scientific theories do not embrace all knowledge and the entirety of the universe; they are restricted in applicability.
Having lanced the boils of justificationism, neojustificationism, and dogmatic falsificationism, Lakatos introduces the shining knight of methodological falsificationism. Before the knight is brought into the arena Lakatos first detours through the thickets of conventionalism. The path is traced through pairs of alternatives, each pair being presented and one selected for further exploration.
Choice the first: Passivist versus activist theories of knowledge
‘Passivists’ hold that true knowledge is Nature’s imprint on a perfectly inert mind: mental activity can only result in bias and distortion. The most influential passivist school is class empiricism. ‘Activists’ hold that we cannot read the book of nature without interpreting it in the light of our expectations or theories. [p20]As an observation, the most influential passivist schools are the various forms of mysticism, for which see Star Wars.
Choice the second: Conservative versus revolutionary activist theories
‘Conservative activists’ hold that we are born with our basic expectations; with them we turn the world into ‘our world’ but must then live for ever in the prison of our world…. But revolutionary activists believe that conceptual frameworks can be developed and also replaced by new, better ones; it is we who create our ‘prisons’ and we can also, critically, demolish them. [p20]Lakatos instances Kant and Kantians as conservative activists and, in a footnote, also Hegel. Lakatos says that Whewell, Poincare, Milhaud and Le Roy opened the revolutionary activist door which now, apparently, acquires the label “conventionalism”.
Choice the third: Conservative versus revolutionary conventionalism
Conservative conventionalists hold that there is a standard sequence of stages. The first stage is a period in which theories are developed by trial and error. The second stage are inductive epochs in which the best theories are ‘proved’ by a priori considerations. The third stage is the cumulative development of auxiliary theories. The upshot is that well established theories are ruled to have been proved by a methodological decision and are not refutable.
Revolutionary conventionalists, on the other hand, hold that theories are not permanent prisons and are always potentially demolishable.
Choice the fourth: Simplicism versus methodological falsificationism
Lakatos identifies two rival schools of revolutionary conventionalism, Duhem’s simplicism and Popper’s methodological falsificationism. The essence of simplicism is that the simpler (more elegant) theory is to be preferred; it is subject to the objection that the criterion is highly subjective and is a matter of transitory fashion.
We have now arrived at Popper’s methodological falsificationism which Lakatos proceeds to explore in more detail.
We may begin with the notion of “unproblematic background knowledge”. Scientific knowledge forms a network of theory and observation which is refined and elaborated over time. All theories and observations are open to question and challenge; none are taken as being absolutely certain. This testing is done in the context of other scientific knowledge which is provisionally assumed to be unproblematic.
The difference between dogmatic and methodological falsificationism is that the former treats experiment and observation as being absolutely reliable as falsifiers whereas the the latter treats them as being provisional. Under methodological falsificationism theories are marked as rejected, i.e., classified as unscientific; under dogmatic falsificationism they are marked as falsified. In other words the shift from dogmatic to methodological falsification is a shift from truth to admissability. Falsification abandons the notion of truth as such and uses the weaker notion of “not known to be false”. Methodological falsification in turn replaces “false” by “rejected”.
Choice the fifth: Naive versus sophisticated methodological falsificationism.
The fundamental difference here is that naive methodological falsificationism rejects theories whereas sophisticated methodological falsificationism replaces. (MF hereafter will stand for methodological falsificationism.) In naive MF theories are rejected when they are “falsified”; in sophisticated MF theories are replaced by better theories.
The difficulty with falsification is that a “falsified” theory can always be rescued by an auxiliary hypotheses. in naive MF the falsification decision becomes one of deciding whether to accept the rescuing hypothesis. In other words the decision must be made as to whether the falsifying “datum” is an anomaly to be provisionally ignored or whether it is a damning crucial experiment.
In sophisticated MF a replacement theory must meet two acceptibility criteria. The first is that it must have additional empirical content, i.e., it must lead to the discovery of novel facts. The second is that the some of this excess content must be verified. In addition it must explain the previous success of the theory it replaces.
Contrary to naive falsificationism, no experiment, experimental report, observation statement, or well-corroborated low-level falsifying hypothesis alone can lead to falsification. There is no falsification before the emergence of a better theory. [p35]Lakatos characterizes the ‘falsification’ situation as needing a pluralistic model. The situation is not of a conflict between theory and facts but rather one between interpretive theories and explanatory theories.
Sophisticated MF, then, leads to the concept of a series of theories, each an improvement on its predecessors. Counter evidence is always relative. That is, counter evidence is evidence that serves as a refutation of the theory being replaced and as confirming evidence for the replacement theory.
The generation of these theories, then, takes place within the context of research programmes.
A research programme will have a negative and a positive heuristic associated with it.
The negative heuristic bars tampering with the ‘hard core’ – anomalies and counter instances are not accepted as refuting the hard core, are not immediately taken as being fatal. Instead the core is rescued by auxiliary hypotheses. A programme remains progressive as long as it continues to increase in empirical content.
The positive heuristic is the research policy of the programme – the puzzles to be solved, the models to be constructed, the questions to be investigated – in short, the substance of Kuhn’s “normal science”. Lakatos emphasizes that methodology of research programmes results in the relative autonomy of theoretical science.
Lakatos raises the question: How do research programmes die? Are their objective criteria for their death or do they simply die as a consequence of changing scientific fashion?
His answer is that programmes are eliminated when superior programmes supersede them. This is, in his view, the rational reason for their death. If a programme ceases to be progressive, i.e., it is no longer generating theories with no empirical content, then it remains a part of the body of science.
Note: programmes do indeed fail for “social” reasons.
In the competition between programmes fledglings are given leeway. New programmes (which are continually being started up and usually are abandoned) are given a chance to establish what they are good for. The real competition is between programmes which start with different aspects of a domain and encroach on each other. Where they conflict experiment may decide between them; then again it may not.
Lakatos says that there are two kinds of crucial experiments – those that decide between theories within a research programme and those that decide between research programmes. The former are part of the normal process of scientific investigation.
Lakatos holds that real crucial experiments – those which establish one programme over another – are seldom recognized at the time as being crucial. What is more, experiments intended to be crucial often are not. The essential difficulty is that it is only after the conflict has been resolved that one can recognize what the experiment signified. Within each research programme the putative crucial experiment is interpreted differently.
It is notable that Lakatos disparages Kuhn and in nowise represents him correctly (or favorably). Thus in the section on crucial experiments we have:
One must never allow a research programme to become a weltanschauung, or a sort of scientific rigor, setting itself up as arbiter between explanation and non-explanation, as mathematical rigor sets itself up as arbiter between proof and non-proof. Unfortunately this is the position which Kuhn tends to advocate: indeed, what he calls ‘normal science’ is nothing but a research program that has achieved monopoly. But, as a matter of fact, research programmes have achieved complete monopoly only rarely and then only for relatively short periods, in spite of the efforts of some Cartesians, Newtonians and Bohrians. The history of science has been and ought to be a history of competing research programmes (or, if you wish, `paradigms’) but it has not been and must not become a succession of periods of normal science: the sooner competition starts, the better for progress. `Theoretical pluralism’ is better than `theoretical monism’: on this point Popper and Feyerabend are right and Kuhn is wrong. [pp 68-69]Now this is quite wrong. Kuhn’s `normal science’ is very much Lakatos’s `positive heuristic’. It is the work that goes on within the context of a research programme as Lakatos grudgingly admits in a footnote [p91]. Kuhn does not advocate theoretical monism. One cannot equate paradigms and research programs although they are intimately related. Both Kuhn and Lakatos make the same mistake on behalf of their intellectual children, Kuhn for his paradigms, Lakatos for his research programs, which is to fail to recognize that they come in varied sizes, scope, and temporal duration. That is, effects due to these variations are not reflected in their discussions.
In the conclusion Lakatos distinguishes between mature science, consisting of research programmes, and immature science which consists of “a mere patched up pattern of trial and error”. Mature science has “heuristic power”.
The paper has quite a few examples – case studies – which illustrate his various arguments; they constitue much of the text. These are well worth reading carefully. The section, Kuhn vs Popper, might profitably have been omitted.
Remarks in commentary
In reviewing a major work such as this one looks not only at what is said but also what is not said – what themes, thoughts, and topics are absent.
One of the striking omissions is a theory of scientific truth. There are fragments of a theory; that is what the progression through the theories of knowledge is about. It is a curious progression; we move through:
TrueIn what sense, then, does scientific knowledge have any truth content at all? In what sense, then, can it even be said to be knowledge? The answer, perhaps, lies in the notion of verisimilitude. Verisimilitude is, however, a notion with serious problems. Lakatos wrestles with the problem here and there, mostly in the footnotes, but comes to no resolution or even serious treatment. In practice discarded theories of scientific truth appear as “unproblematic background knowledge”.
A second omission is the omission of everything except astronomy and physics. One might well ask whether his structure applies to anything except physics (notoriously considered the hard, mature science.) Perhaps it does; perhaps it does not; he does not begin to consider the question.
A third omission is that the question raised in the introduction, “What is the difference between science and pseudoscience” is never actually answered. The main paper [written earlier] distinguishes between mature and immature science and between progressive and degenerating research programmes. The introduction [written later] names various pseudosciences, e.g. Marxism, but `demolishes’ it as having a degenerate research programme. In `Real’ science, however, a degenerate research programme lives on as long as nothing better has come along.
The depiction of research programmes and the march of theories is rather schematic and doesn’t accord well with actual practice. Thus:
He doesn’t really come to terms, in my opinion, with why science works. This need not be counted as fault – it is not the question he was engaged in answering.
This page was last updated January 23, 2002.