One obviously poor and one disputable good objection to normative demarcation criteria

I present one patently poor--but popular--objection to 'normative' demarcation criteria. This objection can be put as follows: there is a disjunct between the normative rule and historical adherence to the rule. It is often (mistakenly) attributed to Imre Lakatos, Paul Feyerabend and Thomas Kuhn. I then show why this objection is misdirected against all proposed solutions to normative problems of demarcation.

I then present the more nuanced objection seen in Lakatos, Feyerabend and (implicitly) Kuhn's writings. This objection was thought to be the death knell of normative demarcation criteria distinct from history of science. The objection runs as follows: there is a disjunct between what we ought to do in the long-term based on these normative criteria and what we ought to do in the short-term based on a desire to avoid falsehoods and preserve truths. The traditional normative demarcation project seen in, for example, Karl Popper, is therefore 'superseded' by the 'historical' turn in Lakatos, et al.

I explain how clarifying the more subtle objection is enough to show that it is overstated. Thus, Popper's solution doesn't succumb to this form of objection, although there are ways to improve Popper's approach--specifically by incorporating insights by Feyerabend, Lakatos and Kuhn. Furthermore, no other proposed ahistorical solution to the normative problem similarly succumbs to similar objections. The traditional ahistorical project isn't superseded by the historical turn per se; rather, the two play different functions in philosophy of science with different areas of focus.

1. The empirical argument against normative demarcation criteria

There is one objection that is often directed at normative demarcation criteria. It runs more or less as follows:

The refuting empirical argument (REA): Certain methodological or normative rules or 'best practices' in the sciences are not, or have not been, followed in scientific practice. These rules stand refuted by the empirical evidence, and therefore should be rejected on those grounds.

An example of this argument can be found in Deutscher (1968). This objection is, as far as I have seen, almost exclusively directed at Popper's normative demarcation criteria, however it applies just as (in)effectively to other proposed normative criteria.

I expect that if more attention is directed at more modern-day normative demarcation criteria proposed in, for example, Pigliucci and Boudry's edited volume, Philosophy of Pseudoscience (2013), we should see a resurgence in this same (very poor) objection in the coming decades.

Upon more than a moment's reflection, it is clear that REA is not a cogent criticism. Here are two reasons why.

First: some (perhaps many) scientists will on occasion follow bad practices. It would make as much sense to argue that since certain normative rules or 'best practices' in policing are not, or have not been, followed by police officers, these rules stand similarly refuted by the empirical evidence. Unlike laws of nature, we can and have violated normative or behaviour-binding epistemic and moral rules. The proper response to this conflict between rule and behaviour is to conclude people have failed to adhere to these rules, not treat these rules as being refuted by the empirical evidence.

Second: it is possible, that despite the apparent successes in the natural sciences, that most scientists have failed to uphold these normative rules. Even then, this would not constitute a refutation of these norms. Although it would be prima facie strange that a norm did not 'square' with even paradigmatic examples of 'best practices', at most it may lead us to suspect the proposed normative demarcation criteria require an additional explanation for continued apparent success in spite of--rather than because of--adherence to the rule.

A corollary of the REA is another argument. Call it CEA:

The confirming empirical argument (CEA): Certain methodological or normative rules or 'best practices' in the sciences are, or have been, followed in scientific practice. These rules stand supported by the empirical evidence, and therefore should be accepted on those grounds.

The first and second reason for rejecting REA are mirrored in the rejection of SEA: even if all scientists behaved in ways that adhered to a normative criterion of demarcation, it would not follow that the norm received empirical support. This result may be, however, a bit more surprising, but only if we fail to realise we have smuggled in the assumption that most scientists behave in ways that we believe are worth emulating. What we believe is worth emulating may be mistaken.

REA and CEA both implicitly include the following two assumptions:

A: Good scientific practice is such-and-such. 

B: Most scientists adhere to good scientific practice.

Without assuming A&B, the entirety of REA and CEA are argumentatively empty. The original objection is revealed to be the following implicit argument:

1. Good scientific practice is such-and-such.

2. Most scientists adhere to good scientific practice.

3. If scientists follow a proposed normative demarcation, scientists will no longer adhere to good practice.

4. Therefore, a proposed normative demarcation criterion at odds with such-and-such conflicts with good scientific practice.

So much for the REA and CEA. 

There is a further puzzling result: if we reject the REA and CEA, no empirical evidence could (at least on its own) support or refute a normative demarcation criterion. But what use is there for history of science if history of science cannot inform attempts at solving the normative demarcation problem? If REA and CEA are rejected, attempts at describing 'normal' scientific practice tell philosophers of science nothing. The entire 'historical' turn seen in Lakatos, Feyerabend and Kuhn would be (presumably) predicated on a mistake.

2. Properly appraising proposed normative demarcation criteria

A good working assumption is that this conclusion is (obviously) mistaken. Some of the best philosophers of science in the 1970s cannot have overlooked these problems with the REA and CEA.

Here's one way to skirt this problem and salvage the plausibility of the 'historical' turn: methodological or normative demarcation criteria don't assert that scientists in fact follow rules of appraising theories or their behaviour. Rather, they specify (a) an aim or group of aims, and (b) a set of rules for satisfying those aims.

It is the job of a philosopher arguing in favour of normative demarcation criteria to specify (c) why this aim is both worthwhile and realisable, as well as (d) provide (plausible) reasons for why these rules will (plausibly) satisfy those aims faster, easier and/or more reliably than (known) alternative rules.

If we are to criticise any proposed normative demarcation criteria, we must criticise (a) and (b) on these grounds--their failure to satisfy either (c) or (d), rather than the REA and CEA.

A proposed normative demarcation criterion that is not realisable would be like an ethical norm that could not be put into practice--it may be worthwhile, but we cannot realise it. That is not to say that we should not still aim at it; rather, some nearby aims may be preferable because they are more realisable, given our epistemic limitations.

This trade-off between whether an aim or group of aims is worthwhile or realisable runs in the other direction as well: perhaps, if our present condition were different, our current group of worthwhile aims can expand to include them as well.

The most plausible objections are raised at how (a) cannot be satisfied by current rules proposed in (b), thus necessitating either the rejection and replacement of (b) with other rules, or how no possible or feasible rules can satisfy (a), necessitating the rejection and replacement of (a). This requires a fully different approach than with the REA and CEA, namely whether (a) is worthwhile and realisable and whether (a) is satisfied faster, easier and/or more reliably by (b) than alternative rules.

As an illustration, it helps to first examine Popper's normative demarcation criteria: they are a consequence of accepting that past predictive success of a theoretical system is not indicative of truth. However, past predictive failure of a theoretical system is, so Popper claims, indicative of falsity somewhere in the theoretical system. If it is impossible to follow a methodological rule that rules in predictive success, such that increasing predictive success will aim at (probably approximately) true theories in the present, we must abandon or weaken our rule. But can we maintain weaker aims? So long as our aim remains directed at increasing truth content and directed away from increasing false content of theoretical systems and predictive failure is indicative of falsity, Popper suggests adopt the strictly weaker aim of minimising falsehood by taking predictive failure to be indicative of falsity of a theoretical system. This is because, although it is incredibly difficult, predictive failure remains indicative of falsity somewhere, even if the falsity should be located in assessments of the experimental setup.

3. Popper's methodological rules

Before continuing on to the criticism leveled by Lakatos, Feyerabend and Kuhn, it helps to cover some of Popper's proposed methodological rules.

(1) Scientists ought to prefer a new theoretical system (research programme, paradigm, ...) that has greater predictive content than its predecessors. 

This rule is an expansion of Popper's (and Ayer and Carnap's) criterion of predictability of statements. We are no longer merely interested in which statements are empirically significant and which are not; now we are interested in which statements have greater empirical content compared to their rivals. The greater the empirical content the better.

(2) Scientists ought to prefer a new theoretical system (research programme, paradigm, ...) that is of greater explanatory breadth and depth of preceding theoretical systems; that is, 'a theory that has been well corroborated can only be superseded by one of a higher level of universality' (2002, 276).

(3) Scientists ought to test a theoretical system as severely as possible (ibid. 281). 

Note that Popper neglected the corollary of this rule which has similarities with argumentative norms: scientists ought to construct theories as robust as possible, that is, the theory that is tested must be the best available version of that theory. We are interested in testing 'mature' scientific theories. 

On this point, Kuhn's work emphasising 'normal' science does improve on Popper's proposed methodological norms, although it is not at odds with Popper's norms. For the life of me, I cannot understand why some critics take the discussion between Kuhn and Popper to amount to more than a dispute over emphasis over the degree of the usefulness of 'normal' science. While it is of use to construct 'mature' scientific theories rather than 'immature' theories, there exist diminishing returns, and it would be a mistake to overemphasise its importance.

(4) Scientists ought to reject theoretical systems that are incompatible with agreed-upon statements about some spatio-temporal location K (i.e. if an epistemic community is in agreement that a theoretical system is incompatible with an observation or experimental result, one of the two must go; if the experimental result is accepted, some part of the theoretical system must go); we can only provisionally accept theoretical systems that have not been falsified. 

That is to say, Popper's much-maligned criterion of falsifiability for theoretical systems, although central to his territorial demarcation, is of far less importance in his normative criteria: sometimes the fault lies with the experimental setup, other times the fault lies with some core assumption, or with an auxiliary hypothesis. There is no way to determine this from the armchair. Sometimes it is a crossed wire; other times it is the entire theoretical scaffolding.

However, Popper suggests that when rejecting the theoretical system [for it must be rejected as a whole], it must be replaced with a 'cousin' or 'sister' system that retains as much of its content as possible. The scientist's aim is to then retain as much content as possible without engaging in mutilation to key theoretical terms, introducing ad hoc explanations for incoherence or failure, and isolating parts of the theoretical system from one another as best as possible in order to check where the problem lies.

(5) Scientists ought not accept inconsistent statements.

(6) Each progressive theory 'contains the old, well corroborated theory--or at least a good approximation to it' (276).

We can include a number of additional methodological norms, such as:

(7) Scientists ought to prefer theoretically and conceptually simpler theoretical systems over their rivals.

This rule is not emphasised as much as the other rules in Logic, although it goes without saying that adopting it helps avoid the objection that predictive content can always be introduced by way of irrelevant conjunction.

4. Finally, the nuanced objection

Continuing this example, Lakatos may disagree with Popper and claim neither past predictive success or failure of theoretical systems is indicative of truth or falsity. If so, Lakatos may change (a) to some related aim nearby to truth, or change (b) such that past predictive success or failure of research programmes is indicative of truth. 

This particular type of criticism was developed by Feyerabend, Lakatos and implicitly by Kuhn: Popper's norms do not satisfy (a), therefore either better norms must replace Popper's or we adopt different (realisable) aims.

It's worth reexamining what evidence Lakatos, Feyerabend and Kuhn presented against Popper's normative criteria. I will not do that here. Rather, I will jump to their conclusion, which amounted to the following: judged purely on a short-term, tactical level, (1)-(7) were violated when examining a number of historical case-studies, and at the time (or in retrospect) we were well-advised to break these rules. In short, all these rules are too severe. In sum, adhering to these proposed rules too rigidly in the short-term goes against our long-term interest of (1). 

Framed this way, much of the work done by Lakatos, Feyerabend, and Kuhn is almost trivially responded to by affirming the following thesis: short-term and long-term interests will almost always conflict in almost any (socially, conceptually, logically) complex, long-term (inter-generational) goal-directed activity.

This thesis, I believe, is true for the following reason: there simply is the constant tension between short-term and long-term interests in almost all social activities. If short-term interests are prioritised over long-term interests in times of war, to take one (unfortunate, but illustrative) example (or with chess), if the aim is to maximise immediate success with short-term tactics (e.g. immediately taking an open position) this will assuredly go against long-term strategy because there is no care taken to see whether taking the position will help achieve some long-term aim (such as winning a game of chess); if long-term interests are prioritised, long-term strategy is bound to clash with short-term tactics, since strategic rules will rigidly direct tactical decisions that are deleterious for some on the tactical level (e.g. a sacrificial rearguard to protect a withdrawal).

Unlike chess (or even war), the natural sciences and other forms of empirical inquiry that involve inter-generational goal-directed activities (e.g. philosophy, politics, the social sciences, history, mathematics, etc.), are attempts at solving some of the most difficult, intractable and intransigent problems. It is natural that these sorts of problems involve some of the most complex, energy-exhaustive and time-intensive attempts at solving them, and are bound to include numerous historical case-studies that conflict with the overarching long-term goals. There will be dead-ends, moments that theory advances beyond available technology and searches for proverbial needles in haystacks.

But if  the superior objection raised by Lakatos et al. cannot undermine Popper's proposed normative rules, what can? I'm writing up some notes on Feyerabend's proposed improvements to Popper's normative rules. They should be posted sometime in the upcoming weeks.


Deutscher, M. (1968). 'Popper's Problem of an Empirical Base'. Australasian Journal of Philosophy 46: 277-288.

Pigliucci, M. & M. Boudry (2013). Philosophy of Pseudoscience: Reconsidering the Demarcation Problem. Chicago: University of Chicago Press.

Popper, K. (1935) [2002]. The Logic of Scientific Discovery. London: Routledge.