Remarks on sophisticated methodological falsificationism

Recall Lakatos’s example of observed planetary ‘misbehaviour’. He says this might be taken as a refutation of the existing Newtonian celestial mechanics but on the other hand it might be taken as evidence of the perturbing effect of another planet. Such a hypothesis suggests possible crucial observations of the postulated planet concerned. But if no such planet is observed, this still need not refute the theory that there is such a planet (itself deployed to defend existing celestial mechanics), which can in turn be resisted by deploying the further hypothesis that there is opaque intervening material. And so on ad infinitum. In other words, no apparently falsifying evidence is ever decisive. Lakatos concludes that: ‘Scientific theories are not only equally unprovable, and equally improbable, but they are also equally undisprovable.’ (p. 103). Rejecting dogmatic falsificationism, Lakatos goes on to consider two further revisions of falsificationism each of which provides insights into the nature of scientific methodology, though neither of which, he argues, sufficiently describes it. The two further versions are:

Naive methodological falsificationism: this introduces the idea that observations and theory have to be considered holistically in science, i.e. as working together in groups rather than as isolated observation theory pairs.

Sophisticated methodological falsificationism: the key idea here is of what Lakatos calls a ‘research programme’, i.e. groups of observations and theories operating over an extended period of time.

Interestingly, Lakatos attributes all three versions of falsificationism (dogmatic, naive methodological, and sophisticated methodological) to Popper. His conclusion is that sophisticated methodological falsificationism gives the ‘best fit’ to what goes on in science. This conclusion, however, has been challenged by both historians and sociologists of science.

Naive methodological falsificationism...

Rejecting dogmatic falsificationism, Lakatos introduces the second of three pictures of science he discusses: naive methodological falsificationism. Lakatos attributes this position to some of the writings of Popper, although he later goes on to argue that Popper subsequently developed a yet more sophisticatedmethodological falsificationism (the third of Lakatos’ pictures of science). Surprisingly (or perhaps not given the antecedent arguments), Lakatos reports that this is a form of conventionalism.

... is a form of conventionalism

This is a surprising concession because it appears to give up a key feature of empiricism: that scientific theories answer to the facts. Conventionalism in the philosophy of science is (in general) the label for a view of scientific theories in which they are taken to be adopted for their usefulness and convenience without any regard for their truth. Think of the convention that everyone should drive on the same side of the road. This is useful but in no sense answers to a Platonic ‘fact’ of motoring and is itself neither true nor false. Conventionalism is thus antirealist about scientific methodology. It is also (though again only in general) antirealist in the metaphysical sense: that is to say, conventionalism denies that it even makes sense to speak of scientific theories being true of the world.

Lakatos’s subscription to conventionalism is not so radical, however. Conventionalism enters Lakatos’s account of methodological falsificationism through two decisions that correspond to the two assumptions of dogmatic falsificationism discussed above. Some statements are deemed observational and thus themselves immune from falsification by decision (p. 106). This is a matter of using successful theories as extensions of our senses by allowing them to go unchallenged when they are used to challenge other theories. Thus, rather than relying on a decisive foundation (whether used to confirm or refute theories), methodological Falsificationism is based instead on ‘piles driven into a swamp’ (p. 108). The foundation of observation is only relatively stable rather than completely grounded. Thus even if ‘falsified’ by statements deemed observational, a theory may still be true. Lakatos mentions as an example of historical rashness the fact that Galileo and his disciples accepted ‘Copernican heliocentric mechanics in spite of the abundance evidence against the rotation of the earth’ (p. 115). In fact it turned out that that ‘evidence’ was misleading and Galileo was right to adopt the heliocentric view. Thus the rational advice—that falsified theories should be rejected—is fallible advice. It may result in the rejection of true theories.

Without a firm empirical basis of incontestable observation statements, the original falsificationist demarcation of science from non-science is undermined. However, Lakatos suggests that a conventionalist reworking of it is still acceptable. Theories will, however, only be ‘falsifiable’ in the light of methodological prescriptions about putting them to the test (cf. p. 111). A theory is scientific if it is ‘falsifiable’, but this now requires a kind of intellectual honesty from scientists to treat it as falsifiable and to test it accordingly as in principle the world can never decisively show that it is false.

Sophisticated methodological falsificationism

Despite the fact that naive methodological falsificationism is an improvement over dogmatic falsification in its attempt to come to terms with the holism implicit in theory testing, Lakatos suggests that it is still an implausible account of actual scientific practice. This leads him to suggest his third falsificationist model of scientific rationality, sophisticated methodological falsificationism. The key development in moving from naive to sophisticated methodological falsificationism is to take the unit of assessment in science not as an individual theory standing in relation to evidence, but as a series of theories grouped within what Lakatos calls a research programme. A research programme (i.e. a series of theories) stands in relation both to the evidence and also to competing research programmes. Assessment is thus explicitly temporally extended.

Within a research programme, one theory can replace another if two conditions are satisfied: 1. the new theory predicts facts disallowed by the former theory while explaining the successes of the former theory, and 2. some of its ‘excess content’ (i.e. the additional facts predicted by the new theory) has passed empirical tests.

Within a research programme, therefore, theories are falsified by other theories in the light of evidence, rather than, as in Popper’s original and simple formulation, by the evidence directly. A research programme, i.e. a series of theories, is said by Lakatos to be progressive if both the above conditions are met and to be degenerating if not. Crucial experiments are thus not just any experiments that produce results that are inconsistent with a research programme (Lakatos calls such results ‘anomalies’), but only those that distinguish between competing theories in a series.

Research programmes: hard cores and protective belts, positive and negative heuristics

A research programme, as envisaged by Lakatos, comprises a hard core of central assumptions that are shared between different theories in a series; and also a ‘protective belt’ of further assumptions. The latter are those further theories and hypotheses that are needed to relate the hard core to observations.

For example, current research on changes in neurotransmitters in the brain in conditions such as depression is a research programme, in Lakatos’ use of the term. There are many and competing particular theories of how this or that change in this or that neurotransmitter might be related to depression. But all these theories share, as one hard core of assumption, the belief that neurotransmitters and mood are related. This assumption is surrounded by a penumbra of other beliefs about, for example, neuroimaging techniques, particular brain chemistry and so on. This means that if, for example, a hypothesis about the connection between a particular mood and neurotransmitter were to fail to gain experimental support it could be explained as simply the wrong choice of mechanism, the failure of the imagining technique or whatever without being taken to cast doubt on the general idea that neurotransmitters and mood are related.

Lakatos then specifies two methodological rules: the positive and negative heuristic. The negative heuristic is the prescription that the hard core should be preserved even in the face of observational anomalies. The positive heuristic determines the kind of changes that should be made when there are anomalies. Thus the negative heuristic determines the continuity between different theories in a research programme in that it ensures constancy of the hard-core assumptions. This is deemed irrefutable by empirical evidence: a more or less conventionalist decision. The positive heuristic—which remains sketchier in Lakatos’s account—concerns how potential anomalies are to be dealt with: what sort of auxiliary assumptions are built into the refutable ‘protective belt’.

We can see both heuristics at work in our example (above) of the current research programme on neurotransmitters and depression. Thus, the negative heuristic protects the programme from evidence that social factors have a role to play in depression: such factors, the negative heuristic specifics, must be understood as being mediated by neurotransmitters. The positive heuristic, in this case, concern, among other things, the general shape of mechanisms that might be appealed to. The idea, for example, that the mind has no physiological correlate and might float free of it will receive short shrift.

The history of science is, according to Lakatos, a history of the development of theories within a larger structure of competing research programmes. Research programmes are the units that progress or degenerate—according to how theories are developed in line with evidence—rather than the individual theories themselves.

Rational reconstructions and the history of science

Lakatos’s approach is thus particularly interesting in the way it takes the simple logical point about the asymmetry of refutation and confirmation of universal claims by single instances, which Popper (1972) highlights, and shows that this does not by itself yield a plausible scientific methodology. In attempting to preserve Popper’s emphasis on the negative use of evidence, Lakatos develops a model of a historically extended process in which groups of theories are assessed holistically, i.e. in relation to each other and to the data, which is itself regarded as fallible and to be assessed in the light of theoretical context. But as a result, the methodology cannot give decisive advice about when a theory should be accepted or rejected. In the main, it is better to support progressive rather than degenerating research programmes. However, exactly when it is appropriate to abandon a degenerating research programme is a matter of uncodifiable judgement. There are no settled criteria for this. A programme could be degenerating for any length of time before again making progress. Lakatos’s philosophical model gave rise in the 1970s to a project of writing rational reconstructions in the history of science. Effectively this was a way of viewing the history of science as exemplifying Lakatos’s model of what is rational in the pursuit of science. Events that fitted the model of rational behaviour for scientists were thought to require no further explanation. Only events that did not fit that model called for further historical or sociological explanation.