The Demarcation Problem: What is science?

Here I present Popper, Kuhn and Lakatos’ accounts of science and analyse their adequacy at solving the demarcation between science and non-science, known as the demarcation problem.

Science is a value-laden term denoting authority. Commanding trust and power owing to its success across many areas of human enterprise, the imprimatur of science guides social policy decisions in healthcare, environment, education and journalism. Pseudoscience, on the other hand, is a term of derogation referring to non-science beliefs and practices that are mistaken or false, and often harmful; typical examples include homeopathy and creationism. Pseudoscience goes beyond fraudulent data, the nature of the doctrine and its methods are falsely pretended to be scientific or to represent the most reliable knowledge available, while often insouciant towards making precise or truth-apt claims (Hansson 2013, 2017; Ladyman 2013). In contrast there is consensus that science is impartial, collaborative and self-critical (Merton 1973).

Laudan (1983) advocated that science/non-science distinctions be excised from our vocabulary as crude rhetoric, in favour of piecemeal evaluation of claims and evidence. However many other philosophers of science still value the term “scientific” and pursue its explication (Pigliucci 2013). One purpose of this is to persuade the scientifically nonliterate to belief and action in areas of public debate. Yet science’s heterogeneity and changing nature makes simple demarcation difficult. The pursuit of demarcation criteria traces back to Aristotle’s Posterior Analytics and grew in the twentieth century. Candidate demarcation criteria have targeted entities such as theories, institutions and individuals while taking forms such as necessary and sufficient conditions, Wittgensteinian family relations formalised with fuzzy logic (Pigliucci 2013) or a cluster approach akin to a Linnaean taxonomy (Mahner 2013). I will now examine three accounts of science that have particularly interested philosophers since the 1960’s.

Falsificationism — Popper

Popper (1959) states falsifiability as the criterion of science and that the objects of demarcation are theories. A hypothesis is falsifiable if there exist some possible observations that would render it false. If all scientific hypotheses imply particular observations then falsification corresponds to the logical case of modus tollens and the practical case of testability by a crucial experiment. The criterion developed from the Logical Positivists, who Popper believed had confounded a criterion of meaning and a criterion of science. Under falsificationism, Popper counted logic and metaphysics as non-science but believed there could be meaningful non-science.

Popper (1974) illustrated falsificationism with the example of motivation from Adlerian psychoanalysis, particularly the role of overcoming the sense of inferiority. Regardless of how a person actually behaves, psychoanalysis can be used to explain the behaviour, yet Adler’s theory makes no predictions or exclusions of particular behaviours. The actions of a man pushing a child into a river can be explained by the man’s need to prove his daring but contradictory actions of a second man jumping in to rescue the child can be explained by the second man’s need to prove his courage. In offering no predictions ahead of the facts, Adlerian psychoanalysis prevents the formulation of any crucial experiments that might serve to falsify it.

In contrast, Popper claims that science advances by bold conjectures and refutation, for example Eddington’s confirmation of General Relativity. Yet boldness is historically relative (Ladyman 2002). While Popper states that a theory is scientific if it is falsifiable, he goes further: good science is a matter of degree whereby the more novel the predictions against the known facts, the better. Popper (1974) supports this by stating that the preference of a daring and highly informative theory over a trivial one is a rational axiom of progress.

However, falsification has several problems as a demarcation criterion. Firstly, it is too weak. Astrology made falsifiable claims during the Renaissance which turned out to be false. Similarly, early Marxism made novel predictions, for example that the first socialist revolution would occur in the most industrially developed countries of Germany, France or Britain. Yet under falsificationism, the scientific status of these two falsified theories ought to be no different than that of falsified sciences such as Newtonian mechanics. Another problem of weak scope is that any statement can be made trivially falsifiable by conjunction with an observation statement, for example ‘The Absolute is lazy and the ammeter will show 50mA when the circuit is closed’.

But falsification is also too strong. Singular existentially quantified statements such as ‘There is at least one white raven’ are not conclusively falsifiable as no finite set of observations could entail that the statement is false. Moreover, exploratory research is ruled unscientific but much of science is of this nature with no explicitly formulated conjectures (Hacking 1983), for instance Rontgen’s discovery of X-rays while investigating Crookes tubes. Some laws such as the Conservation of Energy are so fundamental it is unclear what would count as a falsifying observation. Other laws are unfalsifiable in principle, for instance Newton’s First Law is inherently unfalsifiable within Newton’s corpus as all bodies are constantly affected by the external force of Universal Gravitation (Gholson and Barker 1985).

Most single scientific hypotheses are unfalsifiable in isolation; only in conjunction with multiple auxiliary hypotheses and initial states are falsifiable observation statements produced. For example, Newtonian mechanics alone cannot predict the motion of heavenly bodies as we also require initial conditions such as their masses, velocities etc. Duhem (1954) had earlier noted that when a conjecture is refuted, the scientist has two options available: she can either modify her main conjecture or its auxiliary hypotheses. For example, the orbit of Uranus was anomalous in Newtonian Mechanics. Instead of abandoning the falsified theory as Popper prescribes, Newtonians modified their auxiliary hypotheses, postulating the existence of a hitherto-unobserved planet which led Le Verrier and Adams to discover Neptune.

Popper still acknowledged the important and necessary role of the dogmatic scientist (Popper 1970) and proposed that an ad hoc modification of an auxiliary hypothesis introduced to save a cherished theory is only legitimate when simple, unifying and independently testable (Popper 1974). That is, it must predict hitherto unobserved phenomena. Popper (1974) conceded that simplicity can only be described vaguely. To prevent contrived ad hoc hypotheses, he insisted on a further condition that modified hypotheses must actually pass some new and severe tests. However, in practice a tenacious scientist may indefinitely avoid the refutation of her main theory by iteratively modifying auxiliary hypotheses. For example, if a hypothesised planet could not be found then she could modify the mass, orbit or presence of cosmic dust in her theory, or the specifications of the telescope required for successful observation (Lakatos 1968).

Even exemplar theories suffered falsifying observations from the outset (Lakatos 1973). Galilean heliocentrism was refuted by Brahe noting the absence of stellar parallax, Kaufmann immediately falsified Special Relativity in the Annus Mirabilis (Lakatos 1970) and Newtonian mechanics couldn’t fully account for lunar motion. Scientists routinely adopt and use theories they know are already falsified until a better alternative emerges. Thus falsificationism fails to accurately describe paradigmatic cases of scientific practice as they occurred in the history of science, which some view as a reductio ad absurdum of any candidate demarcation criteria.

One last problem is that in falsificationism the likelihood of any hypothesis being true is zero. This contradicts the intuitive appeal that scientific hypotheses receive confirmation, the accumulation of positive evidence for a hypothesis through past observations. Falsificationism cannot explain the rationality of a scientist adopting a corroborated hypothesis over an already falsified one (Howson and Urbach 1989).

Normal Science and Revolutions — Kuhn

Kuhn (1977) describes science as a lifecycle of cyclical change amongst communities respecting particular values, practices and beliefs. His object of demarcation is the community and the specialisation of their practice. It is characterised by a long period of normal science marked by consensus, then a crisis, and then a revolution where a new period of “normal science” commences.

Normal science is a project of puzzle-solving by a community of intercommunicating specialists following a shared conceptual scheme and practices called the paradigm, or disciplinary matrix, which comprises questions, theories, results, tacit knowledge, instruments and techniques. Kuhn states that only a single paradigm holds hegemony during Normal Science, for example the caloric theory of heat advocated by Laplace (Chalmers 2013). Scientists’ activity is to force nature to fit a paradigm and any failure of compliance is usually a temporary anomaly not a final refutation; doubt is cast only on the ingenuity of the scientist and not on the conceptual tools at her disposal. The paradigm allows time for progress and esoteric work free from any terminal challenges. For example, observation became relevant only comparatively late in the development of Copernicanism. Kuhn can thereby account for ad hoc modifications to preserve a cherished theory where Popper’s account struggled; during Normal Science, every disciplinary matrix inherently resists falsification.

Eventually, sufficient anomalies accumulate resulting in a period of crisis and the ascension of rival theories. A case in point would be the ultra-violet catastrophe and photo-electric effect which ushered in the quantum revolution around the turn of the twentieth century.

In Kuhn’s account, a scientific revolution is an exceptional event of widespread, rapid wholesale change in belief from the old paradigm to the new. Instigated by a breakdown in normal puzzle-solving, the change occurs as a mass social and psychological phenomenon akin to a religious conversion, without any rational decision-procedure. For example, problems besetting Ptolemaic Astronomy were pressing because of the need for calendar reform and Kepler’s Copernicanism was influenced by Sun worship (Kuhn 1996). The revolution is the Kuhnian analogue of a Popperian refutation.

As well as a change in laws and the perception of the world, a marked feature of Kuhnian revolutions relevant to the demarcation problem is that methodological standards also change over time; before Galileo, naked-eye observations were the trusted, gold standard of observation but after the revolution, instrumental observations were privileged (Chalmers 2013).

Kuhn’s theory suffers several problems in providing adequate demarcation criteria. Firstly, although Kuhn’s account gives better correspondence to the history of science than Popper’s, it still suffers incongruities with some paradigmatic cases of scientific practice. Kuhn claims activity is scientific if it corresponds to the contemporary paradigm for its domain. However, the continued use of overthrown Newtonian mechanics or the presence of Kuhn-loss (phenomena better explained in old paradigms, such as the similarity of metals in Phlogiston Theory) pose a problem as to when an overthrown paradigm may still be practised as science.

In normal science there are no multiple concurrent paradigms, whereas in the practice of actual science not in crisis there exist rival paradigms, for example Young’s wave theory and Newton’s corpuscular theory of light in the nineteenth century.

Another problem is that the notion of a sudden, gestalt shift in belief is discordant with instances from the history of science which illustrate, for example, that the Copernican revolution took over a century to attain acceptance (Lakatos 1970). A further limitation of the theory is that Kuhn’s case studies refer to physics and chemistry alone, without broader tests against the history of the life sciences, geology, or, for example, post-World War II era projects of Big Science (Bird 2018). Kuhn doesn’t account for large, rapid progress via revision of theory rather than revolution, for example in Molecular Biology. Kuhn later weakened any claim to accurately describing all of science, saying he had created only a candidate program not a covering theory (Heilbron 1998).

While Kuhn pays good attention to the socio-historical character of science, his account of prescriptive demarcation criteria is indeterminate to the point of inadequacy. He stipulates that rational, subjective judgements with respect to accuracy, consistency scope, simplicity and fruitfulness standardly characterise good science (Kuhn 1977) but by his own admission these are left vague in their specification. Moreover, Kuhn repeatedly denies that demarcation can be reducible to an algorithmic decision-procedure and rejects it as an unattainable ideal. Although Kuhn (1991) denied that psychoanalysis was a science and claimed that some social sciences could not sustain the extended periods of puzzle-solving required for normal science, his descriptions of the disciplinary matrix and value judgements are so vague as to permit forms of pseudoscientific enterprise and so they ultimately provide no practical demarcation criteria.

Methodological Research Programmes — Lakatos

Dissatisfied with the relativistic implications of Kuhn’s (1996) proclamation that there is no greater authority than the assent of the relevant community, Lakatos sought a methodological criterion for theory change that reconciled Popper and Kuhn’s theories.

Lakatos describes science as a persistent, collaborative attempt at research through a series of theories over time called a research programme which maintain some set of irrefutable beliefs and shared practices including falsification. Each theory has a hard core of incorrigible beliefs accompanied by mutable auxiliary hypotheses called the protective belt. For example, geometrical optics, atmospheric refraction and measurement problems were the protective belt for planetary motion differing from predictions of the Newtonian hard core (Lakatos 1970).

The unit of appraisal is not the theory, but rather the research programme within which a series of testable theories are generated. (Godfrey-Smith 2003). To apply the term scientific to one single theory is a category mistake. Lakatos (1970) noted that Popper equivocated between Marxism in two different senses: meant as a particular theory with auxiliary hypotheses and initial conditions, then Marxism is refuted, but as a research programme it is irrefutable.

Falsification, according to Lakatos, can be said to have a historical character; there is no falsification or recognition of crucial experiment before the emergence of a better theory. A research programme is progressive if it is theoretically and empirically progressive — making novel predictions and confirming them, respectively. The former requirement accounts that the Copernican programme was progressive even though it only became empirically progressive through corroborated excess content in 1616. A research programme can make progress even though all its theory has come to be refuted. The progressiveness of a research programme is a matter of degree; those not progressing are degenerative. A research programme is scientific to the extent that it is progressive (Musgrave and Pigden 2016).

With Zahar, Lakatos saw Popper’s notion of novelty as too strong; it would not recognise the progress of General Relativity in predicting the known but hitherto unexplained perihelion of Mercury. Scientific successes are novel predictions of phenomena, including postdiction, not just prediction of novel phenomena (Lakatos 1970). Lakatos stipulates that the novel predictions of theoretically progressive programmes must be in a natural and not a contrived manner. Examples of the former include the prediction of Halley’s comet with Newtonian mechanics and of Balmer’s formula with the Bohr model, whereas Ptolemaic epicycles explaining retrograde motion are typical of the latter.

The positive heuristic specifies what scientists should do — how to supplement the hardcore and modify the protective belt to yield explanations and predictions. The advance of a programme involves not only the addition of auxiliary hypotheses but also the development of adequate experimental and mathematical techniques. For example, Newton developed his inverse-square law by first considering idealised point masses. He slowly developed complexity, adding attraction first between sun and planet, then between other planets, then sphericity, then spin etc (Chalmers 2013).

A scientific revolution occurs when a hegemonic, degenerating programme is superseded by a progressive one, though its rivals may persist. Contrary to Kuhn’s idea that normal science is dominated by a single paradigm, Lakatos accurately claimed that the history of science typically consists of competing research programmes.

However, this theory has several problems as a solution to the demarcation problem. Programmes are only ever more or less scientific and degenerative programmes may perform a comeback. For instance, Copernicanism was degenerative in its first 100 years but went on to digest its anomalies. Concerned with stemming “intellectual pollution”, Lakatos’ initial aim was to “provide universal conditions under which a theory is scientific” (Lakatos 1970). Yet his theory is ultimately only of use to historiographers of science in appraising theories from hindsight; it is impotent to advise contemporary working scientists. Feyeraband described this non-prescriptive criterion of science as “Anything goes”, that any proposed methodological rule can be found to have been broken in the history of scientific progress (Feyeraband 1978). Despite the focus on a rational logico-methodological criterion of demarcation in the form of progressiveness, Lakatos conceded that his theory evaluates progressiveness without prescribing that a programme should ever be abandoned. Lakatos accommodates the complexity of a sophisticated falsificationism at the cost of a sharp or practicable demarcation criterion.

It is unclear to what extent scientists must explicitly identify their hard cores for Lakatos’ theory to hold true. In practice, ideas can move in and out of the hard core and there is fruitful exchange between rival programmes. For example, exchange occurred between 19th century aether-based theories of light and Carnot’s caloric theory was absorbed by a rival research programme that saw heat as the motion of matter (Godfrey-Smith 2003). The history of science contains many examples of mutable hard cores: some Aristotelians abandoned the doctrine that motion in a void is impossible, some Newtonians abandoned inertial mass (Laudan 1977), some Newtonians modified the inverse square law of universal gravitation to account for the motion of Mercury, and some Copernicans moved the sun off-centre (Chalmers 2013).

Furthermore, a theory’s conceptual defects such as circularity or inconsistency with established notions will impede its acceptance. For example, Maxwell introduced a new concept of force that conflicted with the accepted concept in Newtonian mechanics. Maxwell’s programme was not accepted in preference to a degenerating rival until after 1887, even though it had been progressive for three decades (Gholson and Barker 1985).

Another limitation to note is that Lakatos takes case studies only from physics from the last 300 years and assumes that we can extrapolate to the other sciences. However, some sciences possess features uniquely absent from physics. For example, the complexity found in living systems or the feedback mechanism that societal knowledge of economic theories influences the marketplace behaviour under study by economics (Chalmers 2013).

Conclusion

I have examined the prescriptiveness and fit with actual scientific practice for Popper’s falsificationism, Kuhn’s normal and revolutionary science and Lakatos’s methodological research programmes in the context of the demarcation problem. Popper’s theory appears sharp at first blush but hides ambiguity in its complex application such as defining auxiliary hypotheses. Overly stringent, falsificationism also departs from describing paradigmatic cases of science, for instance early Copernicanism. Kuhn’s Normal and Revolutionary Science provides a compelling alternative to Popper, accounting for the resistance of nascent science to early refutation. However, the history of science shows that theory change may be neither quick, nor unanimous, nor wholesale. Kuhn offers an improved focus on the socio-historical nature of science but this comes at the cost of normative and methodological vagueness in applying his value judgements and defining disciplinary matrices. Lakatos provides some reconciliation of Popper and Kuhn’s theories while aiming for a methodological account of demarcation. Although the notion of a well-defined, continuous hard core is incongruent with examples of scientific practice, progressiveness is a more sophisticated form of falsification accounting for the novel postdiction of phenomena such as the perihelion of Mercury. Nevertheless, Lakatos’ explicit, broad tolerance for degenerative programmes ultimately offers no practicable or adequate solution to the demarcation problem.

References

Bird, A (2018). Thomas Kuhn (SEP)

Available at: https://plato.stanford.edu/entries/thomas-kuhn/ (Accessed: 10th October 2019)

Chalmers, A. F (2013). What is this thing called science? 4th edn, Indianapolis and Cambridge: Hackett Publishing Inc.

Duhem, P (1954). The Aim and Structure of Physical Theory, Princeton: Princeton University Press

Feyeraband, P (1978). Against Method, London: Verso

Gholson, B and Barker, P (1985). Kuhn, Lakatos, and Laudan: Applications in the History of Physics and Psychology, American Psychologist, 40(7), pp. 755–769

Godfrey-Smith, P (2003). Theory and Reality, Chicago: University of Chicago Press

Hacking, I (1983). Representing and Intervening: Introductory Topics in the Philosophy of Natural Science, Cambridge; New York: Cambridge University Press

Hansson, S. O (2013). Defining Pseudoscience and Science, In Pigliucci, M and Boudry, M(Eds.), Philosophy of Pseudoscience: Reconsidering the Demarcation Problem, Chicago and London: University of Chicago Press, pp.61–77

Hansson, S. O (2017). Science and Pseudo-Science (SEP)

Available at: https://plato.stanford.edu/entries/pseudo-science/ (Accessed: 10th October 2019)

Heilbron, J. L (1998). Thomas Samuel Kuhn, In Isis Vol. 89 №3, pp. 505–515

Howson, C and Urbach, P (1989). Scientific Reasoning: The Bayesian Approach, Chicago: Open Court Publishing

Kuhn, T. S (1977). Objectivity, Value Judgment, and Theory Choice, In The Essential Tension: Selected Studies in Scientific Tradition and Change, Chicago: University of Chicago Press, pp. 320–39

Kuhn, T. S (1991). The Natural and the Human Sciences, In Hiley, D. R, Bohman, J and Shusterman, R (Eds), The Interpretive Turn: Philosophy, Science, Culture. Ithaca: Cornell University Press, pp. 17–24

Kuhn, T. S (1996). The Structure of Scientific Revolutions, 3rd edn, Chicago: University of Chicago Press

Ladyman, J (2002). Understanding Philosophy of Science, London and New York: Routledge

Ladyman, J (2013). Toward a Demarcation of Science from Pseudoscience, In Pigliucci, M and Boudry, M (Eds), Philosophy of Pseudoscience: Reconsidering the Demarcation Problem, Chicago and London: University of Chicago Press, pp.45–59

Lakatos, I (1968). Criticism and the Methodology of Scientific Research Programmes, In Proceedings of the Aristotelian Society, New Series, Vol. 69 (1968–1969), pp. 149–186

Lakatos, I (1970). Falsification and the Methodology of Scientific Research Programmes, In Worrall, J and Currie, G (Eds), The Methodology of Scientific Research Programmes, Philosophical Papers Volume 1 (1978), Cambridge: Cambridge University Press, pp.8–101

Lakatos, I (1973). Science and Pseudoscience

Available at: http://www.lse.ac.uk/philosophy/science-and-pseudoscience-overview-and-transcript/ (Accessed: 10th October 2019)

Laudan, L (1977). Progress and Its Problems: Toward a Theory of Scientific Growth, Berkeley and Los Angeles: The University of California Press

Laudan, L (1983). The Demise of the Demarcation Problem, In Cohen, R. S and Laudan, L (Eds), Physics, Philosophy and Psychoanalysis: Essays in Honor of Adolf Grünbaum, Reidel, pp. 111–127

Mahner, M (2013). Science and Pseudoscience: How to Demarcate after the (Alleged) Demise of the Demarcation Problem, In Pigliucci, M and Boudry, M (Eds), Philosophy of Pseudoscience: Reconsidering the Demarcation Problem, Chicago and London: University of Chicago Press, pp.29–43

Merton, R. K (1973). The Sociology Of Science: Theoretical and Empirical Investigations, Chicago: University of Chicago Press

Musgrave, A and Pigden, C (2016). Imre Lakatos (SEP)

Available at: https://plato.stanford.edu/entries/lakatos/ (Accessed: 10th October 2019)

Pigliucci, M (2013). The Demarcation Problem: A (Belated) Response to Laudan, In Pigliucci, M and Boudry, M (Eds), Philosophy of Pseudoscience: Reconsidering the Demarcation Problem, Chicago and London: University of Chicago Press, pp.9–27

Popper, K. R (1959). The Logic of Scientific Discovery, New York: Harper & Row

Popper, K. R (1970). Normal Science and its Dangers, In Criticism and the Growth of Knowledge, Lakatos, I and Musgrave, A (Eds), Cambridge: Cambridge University Press, pp. 51–58

Popper, K. R (1974). Conjectures and Refutations, 5th edn, London: Routledge

See also:
Knowledge