If you would like to download a pdf version of this essay, click here.

Abstract: Normative methodologists engage in a non-systematic, activity, which is a high-level part of their investigation of nature. They argue for changes or innovations in the methods of physicists. They are typically physicists.
Descriptive methodologists engage in systematic, activity, aiming to codify, and assess the extent of justification, of the current methods regarded by physicists as justified - methods which are the fruits of centuries of work by normative methodologists. They are typically philosophers.
The rationality of the methods that the descriptive methodologist abstracts is ultimately determined either by the consensus of current physicists (descriptive methodology) or by theoretical arguments (normative methodology), not by the historical behaviour of physicists. History may provide illustrations, but not tests; it may be useful heuristically to the descriptive methodologist, as evidence of the past consensus, and - by discovering present attitudes to historical stories - evidence of the present one; but it does not determine the correctness of her methodological system.
Philosophers, confused as to their own aims, have involved historians and sociologists in their confusion.
 

Our Problem:  What are the appropriate methods for finding the best methods for investigating nature?

What is the relationship between the methodology of science, science, the history of science?

   Many kinds of causal threads link the tangled mass of historical events. None of them intrinsically attract our attention, by their importance; none of them glow. Each kind can be highlighted by a torch of appropriate colour. The methodologist's torch picks out the threads motivated by the search for Truth. It is the light of her own torch which she sees glowing, reflected back by the threads of history.
   The historian's and the sociologist's torches illuminate all the threads.
 
 

   Methodologists choose to highlight investigations with the aim of {Truth}; they study methods for achieving it, results obtained using these methods, and the extent of justification of these methods and results. They divide into:
(i) Normative methodologists, typically, physicists, who engage in a normative, non-systematic, activity, which is a high-level part of their investigation of nature. They argue for changes or innovations in the methods of physicists.
(ii) Descriptive methodologists, typically philosophers of science, who primarily engage in a descriptive, systematic, activity, aiming to systematise, and assess the extent of justification, of the current methods regarded by physicists as justified - methods which are the fruits of centuries of work by normative methodologists. Once complete, it can, secondarily, become cautiously active and normative.
   The methods which achieve the aim of {Truth} locate the factors in historical episodes that we call 'internal'. Other aims, other methods, other factors, are called 'external'.

Normative Methodologists :
   Normative methodologists are investigative physicists; their results are a living part of the hypothetical structure of physics, though distinct from the structure of physical hypotheses concerning nature; their results are not about nature, but about the best ways nature can be investigated. While descriptive methodology is about physics, outside physics, looking in, reflecting on physics practice, the normative methodologist does not reflect tranquilly on good practice. She is in the thick of it. Her problem is what to recommend now. She is suggesting how physics research in 1997 should be done.
   In particular, she is not saying how it was done in 1985, or in 1785. She is, however, saying that if her methods are different from the previous consensus, she has reasons why her methods are better than theirs.
   Methodology is a branch of philosophy, and also a branch of physics. This is not surprising - physics was once called 'Natural Philosophy'. Seeking - even just presenting in review articles - the best methods for the investigation of our experiences is a natural adjunct to using these methods to seek the truth. It is not philosophy of physics, but rather philosophy in physics.
   Success for the normative methodologist is seeing, and justifying, the need for a change in the methods Physicists have been using in a certain situation. Here are some great examples: suggesting that we should build up slowly from observation, not leap from it to grand abstract principles; suggesting the method of similarity and difference to obtain hints on causes; suggesting that physicists should devise successive approximations to real events, working with idealised models which may not exactly apply to any real event; suggesting that physicists can reasonably devise grand mathematical hypotheses which are tested only by their indirect consequences; suggesting that physicists can rationally devise realistic theories (eg. for light) far from direct testing, whose truth-content is demonstrated by their having true consequences which were otherwise very low-chance.
 

   The methods of the descriptive methodologist should be similar to those of physicists, since both are seeking the truth; this is self-consistency, achieved by successive approximation, not vicious circularity. She forms hypotheses on the rational/IP methods which are used by physicists, and investigates the extent to which they are justified, given the aim of {Truth}. The primary source of evidence for her hypotheses, analogous to physical data, is the current consensus amongst physicists as to what should be labelled rational behaviour in certain situations. She can seek this by studying:
(i) direct data on what physicists think
(ii) indirect data on what they think, via the recent and distant historical behaviour of physicists; initially unlabelled as 'rational', it is labelled by a contemporary consensus of physicists; it is therefore no more than a story, real or fictional, to illustrate the consensus view; it is an indirect version of (i)
(iii) indirect data via intuitive feelings for what is reasonable; these, on the part of the descriptive methodologist, are a vague, indirect, version of (i), but are convenient for quick reference.
   The descriptive methodologist should initially accept the current consensus labelling. A methodologist who criticises it has passed beyond descriptive systematising, and has become dormative, bidding to join the community of investigators.
   Her situation thus differs in one important respect from that of the physicist investigating nature. Their data comes labelled as 'true' and 'false', at least roughly, while hers comes without a label 'rational' and 'irrational', or internal and external. She has to stick the label onto the behaviour. Historical labels of 'rational' are unreliable. Evidence from history indicates that physicists have rarely acted purely out of the aim of {Truth}. So, in sticking the labels on, she will be initially following the judgements of the present physics consensus.

   One result of descriptive methodology is that the aim of {Truth} underdetermines physicists' decisions, even in apparently typical internal situations. Other factors, such as social ones, flow in to fill the methodological vacuum; the involvement of these external factors is perfectly acceptable, as long as the resulting propositions are not given truth-credit beyond what was justified internally.
   The two extreme methodologies for the descriptive methodologist are, roughly, rationalism and empiricism.
(i) At the rationalist extreme, she could glance at a few cases - maybe classical mechanics going to special relativity - and then, using this as a hint, set about creating a full beautiful theory to explain why there is progress in this case. She could invent special idealised cases to try her theory out. She could then spend time trying to solve logical problems that arise as she tries to set her theory up. She could assume that if it works for this case, it will work for all the others.
(ii) At the empiricist extreme, she could plunge into the messy details of what physicists have actually decided in these hundreds of cases, looking for possible general patterns, natural classifications, rough generalisations, possible laws. On this Galilean step-by-step approach, the grand abstract theory is unlikely to emerge until a late stage in the process. She should try to understand the simplest cases of consensus choice between propositions first. Choices between grand systems, such as the choice between classical and relativistic mechanics, will only be understood much later.
   We can see that the criticisms directed, within physics, at the Paris Occamists by the Italian school at the end of the Mediaeval period, and later at Galileo, could easily be directed at the empiricist approach to descriptive methodology. It is trivial: it is inelegant: it is uncivilised: it is demeaning for refined minds to mess around with all the details when we could be unfolding the essence behind the superficialities.
   The best approach is one in which we regularly refer backwards and forwards between the developing model and the practice. We do not deny the need for hypotheses to aid us, any more than Bacon did, but we continually refer back to the practice. This is what we did in devising the methods that are listed in our essay The Methods Of Physics.

   See the chapter on the methods of physics. The methods look like a synthesis of methods proposed by philosophers of science over the last few centuries. Unlike recent philosophy, it has an encyclopaedic quality, since different methods apply in the various abstracted situations. The justifications for the methods is various: the foundational methods are justified by verbal consistency; some methods can only be justified if apparently unanswerable sceptical doubts are quarantined (which is not a problem, since all investigators ignore these doubts anyway); other higher methods are justified by reference to the history of success and failure of types of investigation over centuries (relying on meta-inductions); others are justified by reference to hypotheses concerning the psychology and sociology of human beings, which have themselves been justified using the foundational methods.

   It does not matter whether it does. It includes:
(i) the theoretical, normative, part of the dispute between realists and anti-realists. If one system could convincingly argue that an investigation based on it is the only one that is rational, then it would be the recommended system. (Whether physicists actually have used one system, another, or both, at various time, is a separate, empirical, descriptive, question, in the history of physics) (See my essay on 'Realism and Anti-realism ')
(ii) analysis of the logical structure of physical hypotheses; giving an account of explanations, basic statements, laws, and theories. This is background work, on which the methodology depends
(iii) the paradoxes of confirmation, (ravens and grue), which are theoretical problems in the justification of certain descriptive methodological systems - of a logical empiricist type; these systems are based, for example, on a certain kind of analysis
(iv) the problem of Induction and sceptical doubt; these are problems in the justification of various methods, some concerning the justification of generalising from particular observations to laws, others the truth-credit of theories purportedly supplied by successful novel fact prediction, and improbable consilience of inductions
(v) the Duhem-Quine problem: can the methodologist provide any kind of justification for choosing to alter one part of a structure of hypotheses rather than another, when it has become inconsistent as a result of experimental work
(vi) Idealisations and the ceteris paribus clause: Can a physicist justify, in a certain situation, proposing a hypothesis with an associated catch-all clause which in principle means that any countervailing evidence can be avoided?
(vii) Is there some simple central principle which encapsulates the scientific method? If there is, it should become visible within the systematised methodology
(viii) Do the results of physics progress? Do they gain more truth-credit? Since physicists claim that they do, the methodological system must attempt to justify the claim
(ix) Can physicists justify abandoning a theory after it has had a certain number of ad hoc changes?
   Furthermore, the classic simplified versions of scientific method - Inductivism, Logical Positivism, Logical Empiricism, Bayesianism, Rationalism, Hypothetico-deductivism, Naive Falsificationism, Sophisticated Falsificationism, Scientific Research Programmes, Normal and Revolutionary Science, and so on - can all be naturally regarded as descriptive methodological systems.
   We are trying to avoid speculation on such factual questions as "Are the list of aims of the methodologist actually the aims of present, or past, philosophers of physics?". The only way of answering this would be to collect evidence on the aims of all people called 'philosophers of physics'. We are not interested in this particular question, though other people may be.
Instead we are proposing certain sets of aims which some people might value, and which would characterise certain activities.
   We assume that no-one will ask the question: "What is the real , essential, aim of philosophy of physics? (or, indeed, of history, or sociology)".

   Descriptive methodology, tied, as it largely is, to the requirement that it must fit with evidence of physicists' decisions pre-labelled as rational (intuitively, or by consensus - at the time, or now) will tend to reflect change, but not initiate it. Normative methodology, however, is an independent part of our investigative system; it will initiate change as a result of new arguments, or new evidence of what works.

   The question is, perhaps, whether this is this the same as Quine's theory - and hence either not very interesting or inclined to suffer from the same problems as his theory.

(i) As far as I understand it, one way that an approach can be called 'naturalised' is if it involves the attitude that the natural psychological tendencies of people can be used as justifications - or at least as explanations - of the methods they use. For example, the method of generalising from limited experience could be ascribed to custom and habit, a product of our genetic makeup, itself determined by evolution. This is an explanation, but is not a justification. Perhaps the statement that "There are no fundamental justifications of these methods, only descriptions of what people naturally do" would describe some aspect of it. If so, this is not the view we are taking, except perhaps in the case of the Inductive Presupposition. Nonetheless, the later writing of Quine does seem to have distinct similarities with our view.
(ii) We are not using the results of science - say, of psychology - to establish our foundational methods. If we do refer to such results to establish high methods, we assume that they have been established using the foundational methods.
(iii) The foundational methods of our systematic methodology are not absolute principles of rationality independent of human aims; we suggest that a conventional analysis of rationality indicates that no such principles exist. The aim of {Truth} is chosen by people, specially valued by some people some of the time. When a person does not give it priority, he can rationally ignore our methods - hard though this may be for an intellectual to accept. In this respect, a descriptive methodological system is linked to a particular natural interest of a group of people.
 

   ON THE ONE HAND : Normative investigative methodology, a part of the uncodified intuitive apparatus of a physicist, is an independent part of humanity's search for Truth; it is not limited by history, and is able to progress by argument and evidence.
   BUT ON THE OTHER HAND : Descriptive methodology, the philosophers' attempt to codify intuitive investigative methodology, primarily a reflective task, initially should rely completely on the correctness of the contemporary intuitions of the consensus of investigators.
   Thus when the philosopher looks at history, the judgement of a particular physicist, say Thomson, on whether what he did was justified at the time, should not be taken as decisive. Nor should what the consensus judged at the time. The view of a contemporary historian is equally indecisive. But what the physics consensus judges now should be taken as decisive.
Whatever complicated set of reasons actually explain historically why physicists preferred the wave theory of optics to the particle theory is irrelevant to the question "What rational reasons were there then - or are there now - for preferring the wave theory, given the aim of seeking the truth about Nature?".

   For normative and descriptive methodology, the present methods used in physics, the system presently recommended for use, is the only one that matters. That state is the culmination of centuries of experience, but is unlikely to be its final state.
   The methodological system in physics is in a position far from the direct influence of experience, like a high-level physical theory. Therefore observation or experiment are unlikely to change it suddenly, from below. Evidence could accumulate to suggest that human nature has been wrongly summarised, or that institutions do not work as we had expected, but I am not aware of any changes driven by such evidence. But very important changes have come from below, when failure of the overall investigation has led to a rethink of, say, an entire strategy for devising and testing low-observability claims. Some major innovations over the last 400 years have already been listed.
   Its position makes it open to influence from above, from philosophical theories. This can cause changes, as shown by the effect of positivist philosophy.
 

Simplicist Methodology : This activity is based on the presupposition that there is a simple key feature to be found, by which the methodological system can be unified. Alternatively it includes the aim of finding such a feature. Either way, we do not think that it is rational, unless the desire for simplicity is placed above the desire for truth.
   We cannot justify the use of simplicity considerations when (i) devising methodological theories (ii) justifying preference for methodological theories (it is difficult enough in physics). We have no reason, based on the history of methodology, to expect the methods to form a simple system, unified by some few principles. It seems more likely - from experience of human behaviour - that a true system will be a heap of disorganised methods appropriate to various situations. The simplicity of a proposed system is therefore an indication not of probable truth, but of falsity. The simplicist research programme seems to have been steadily degenerating, rather than progressing - despite epic efforts by a succession of philosophers. Disappointingly, this form of methodology should be judged irrational.
   Worse still, our guiding theory, based on evidence, that human behaviour, even in search of a single goal, tends not to be describable with a simple set of methods, leads us to be suspicious of the truth of any model which is simple. This is painful to us; but we should not let our desire for unity, simple explanations, tidy patterns, distract us from the requirements of our overriding aim - to find out the truth about our best systems for the investigation of nature.
 

Complete-justification Methodology : This special activity modifies the aim to:
(iii) Seek methods which will assist, beyond all doubt, in achieving the aim of {Truth}, in various situations.
   This unquarantined methodology has always interested philosophers, and not interested physicists. Thousands of years of thinking have failed to find the desired methods, which begins to indicate that they do not exist. Once again, this research programme seems to have been steadily degenerating. Disappointingly, continuing to work within it is therefore irrational.
   A variant on this is to presuppose that the methods are fully justified. The key problem then is to find the justification. Any philosophers, fascinated with the problem of sceptical doubt and the justification of inductive methods, who have tipped over into this form of investigation are behaving irrationally.

   The next form of methodology is not irrational, but it is unnecessarily restricted in scope:

Focussed Methodology : Practitioners of this special activity focus their attention on just a couple of situations in physics, which they feel are especially important. For example, a methodologist could be interested only in the methods currently available for choosing between high-level theories, the extent of their justification, and the extent of reliability of the resulting choice.
   This sounds harmless. But it leads to a loss of perspective, in particular on the relative importance of these situations.

   The normative methodologist is presumed to be thoroughly familiar with contemporary physics investigatory methods, the justifications usually claimed for them, the consensus on what present and past decisions seem intuitively rational, and the kinds of situations in which physicists find themselves. As we have seen, such a thinker aims, by careful thought and persuasion, to both criticise and change the methodological attitudes of physicists.
Since she is independent of any particular investigation, she can attempt to demarcate investigations, assessing the extent of rationality they display, relative to the aim of truth.

   A philosophically-trained methodologist typically starts with the disadvantages that:
(i) she lacks this intuitive familiarity, due to probable compartmentalisation of studies
(ii) she has certain special interests, characteristic of philosophy, such as defending claims from sceptical doubt, which are not shared by physicists.
   However, she has the advantage of time to spare, because she is not seeking truths about nature; and she is interested in these matters. She cannot realistically hope, at least initially, to affect the current methodology used by investigators, but she can reflect on what they do, and to systematise it. His role therefore will almost certainly be descriptive rather than normative.
Consider a musical analogy, as illustration: Robert Schumann hardly wrote anything systematic about the methods of musical composition, yet radically altered the methods used; Rimsky-Korsakov, who composed music less famous, wrote a book systematising the methods of orchestration, including his own successful innovations; Walter Piston, who had very little success writing music, wrote a series of books on the classic methods of counterpoint, harmony, and orchestration, containing no critical suggestions of his own (which would border on impudence) but being excellent systematisations of the accepted methods. We suggest that philosophers of science are in the situation of Piston - except that they have usually not written any music at all.
   We should therefore rely completely on the contemporary intuitions of the practitioners as to good internal practice. If our systematisation leads us to assess behaviour, in recent or distant History, differently from the current consensus of physicists, then we should humbly assume that we have got it wrong.
 

(i) Physicists are not always taught what the aim and methods of their profession are - they seem to be expected to pick them up as they go along. They are systematically taught the content. Perhaps teaching the aim and methods might lead to more progress towards the aim. For example, some physicists might not have consciously realised that some methods are important, and might stop using them as a result - peer-referenced journals could be undercut by the Internet.
(ii) An activity is only rational if its aim is truly desired above the other aims of the practitioners. {Truth} may be so valued; but perhaps other human values, such as {Solving Human Problems}, are desired more, and perhaps {Truth} is not always the most promising way of achieving this. Do we want to discover the truth about genetic links between sex, race, colour, and intelligence? Might we not live more happily if, in some areas, we ignored the truth and lived instead with myths?
(iii) The methods judged to be the best at any one time have changed, and will probably change again. Therefore investigators should be consciously aware of the methods, and their justification, so that they can change the methods if circumstances suggest that they are becoming a handicap.
(iv) We can hope to assess the extent to which physicists, and other groups of people who share the same aim, use our methods. If groups are not using the methods, and yet claim to be achieving {Truth} nonetheless, we can criticise them.
(v) Since investigators in high-rated activities may not be consciously aware of their aims and methods, and how they are justified, they may be at a loss when faced with criticism of their activity from outside. (The rather tentative and inconclusive response to creationism could be an example)
(vi) By understanding the philosophical justification, warts and all, for their methods, physicists would understand the extent of constraints on them, and the freedom that they have, within the aim. Conjectural realism, for example, would be seen to be an inescapable implication of the aim (as Einstein insisted). The understanding would partially protect them from the distortions that either hearing about bits of philosophy, or thinking up bits for themselves, may lead to.

   Philosophers of science cannot hope to make progress step-by-step towards an aim if they are unclear as to their methodology - if their aim, and methods, are not agreed. We propose that descriptive methodology, and normative methodology, are two rational methodologies. They have achievable aims, and justifiable methods.  They do not make presuppositions which clash with the evidence, in the way that simplicist and complete justificationist methodologies do.
   Normative methodology should remain the province of scientists, but descriptive methodology is a worthwhile task for philosophers.
   Historical and sociological evidence provides hints to the methodologist in the context of discovery, but is valueless in the context of justification.

Back to philosophy of physics map page