A far better objection to normative demarcation criteria

These are follow-up notes on my previous post on how to present plausible objections to normative demarcation criteria. In it, I present a brief overview of a problem related to the value problem in epistemology, then illustrate two ways of framing any 'final value' to normative demarcation criteria by examining 'early' and 'later' Karl Popper's views on the subject. I end by presenting the mature objection raised by Paul Feyerabend that targets both views held by 'early' and 'later' Popper.

1. Which values are appropriate?

Here is a far better objection than the REA and CEA, and the objection raised by Lakatos, Feyerabend and Kuhn. It begins with the following question: why accept Popper's normative demarcation criteria set out in (1)-(7)? For that matter, why accept any methodological principle? There are a number of potential answers. For example, it may be because they are...

(a) valuable in-themselves. There is no other rationale: scientists ought to put forward theoretical systems of ever-increasing coherent empirical content and unifying explanatory breadth and depth because scientists and philosophers of science agree that is a desirable aim.

(b) there are non-truth-related grounds for why we should adopt them. It may be socially or ethically desirable to produce explanations that unify disparate fields or set out the limits of what is possible.

(c) there is some epistemically valuable reason that isn't aimed at truth, nor seen as a derivative aim of the pursuit of truth, but is nevertheless desirable. That is, ever-increasing empirical content, simplicity and universality of explanations may be epistemically valuable even if these explanations do not aim at truth.

(d) there is some truth-related epistemically valuable reason: these rules aim at empirical adequacy, truth, truthlikeness, aim away from falsehood, etc.

This is a problem of determining whether the grounds for demarcation rules are suitable. (a) terminates without any further grounds, while (b) makes the problem of demarcation no longer relevant to epistemology or philosophy of science.

While it certainly would be a greater benefit were (b) to be true in addition to other grounds, neither (a) nor (b) seem particularly desirable on their own. If (b) were accepted and it was a matter of mere convention that we accept these rules, we would desire something over and above mere convention--convention alone does not entail obligation to uphold convention. Therefore, our sights turn to (c) and (d) as plausible candidates.

Only if (c) and (d) are exhausted and found wanting as plausible grounds should we determine whether (a) and (b) are true, and hopefully the available grounds remain normative, such as finally terminating in an ethical or argumentative imperative to behave in certain ways and refrain from behaving in other ways.

(c) may fare no better: it seems little more than a restatement of (a), with the added benefit that these grounds are somehow valuable for an additional reason--they are not merely valuable in-themselves, but valuable because they are valuable in-themselves and this value is considered a core epistemic principle (although unrelated to truth).

This issue shows that grounding demarcation criteria produces an issue related to--but fully distinct from--the value problem. The earlier Meno problem helps sketch out the value problem. It is as follows: if knowledge is distinctively more valuable than mere true belief, what is that value over and above simply believing truly? As an illustration, it could be that, as Socrates claimed, knowledge is 'tied-down' to the truth, in a way that prevents the easy loss of true beliefs, but any number of dogmatically-held true beliefs that fall short of knowledge may be 'tied-down' to the truth just as strongly.

While the value problem is related to explicating the final value of knowledge that is over and above the value of mere true belief, demarcationists are left with the problem of explicating the final value of certain methodological norms. Is it (a), (b), (c) or (d) (or some combination of the above)? Is there some final value (or should we act as if there is some final value?), or should demarcationists be value-pluralists, accepting more than one final value of demarcation criteria?

We therefore have the following question for philosophers of science: why should we believe unifying theories with immense explanatory depth and breadth are valuable?

There is, furthermore, a second question for the demarcationist: assume we have answered the first problem and have accepted some final value(s) for proposed demarcation rules. Since the REA and CEA are not permissible objections to demarcation rules and the historical objection raised by Lakatos et al. cannot lead to revising normative demarcation rules, there is no obvious empirical evidence available to check whether implementing a proposed rule will reliably lead to successfully attaining the desired final value(s). Why should we believe, for example, that testing theories will, in the long run, direct us towards the truth or away from error? We could operate on the assumption that adhering to the rule is more reliable than not testing our theories--however, that doesn't resolve this problem. We could very well operate on a different assumption that adhering to a different rule is preferable because it is more reliable than other approaches. Which rule is to be preferred, then, and on what grounds?

2. What are these norms or rules aimed at?

It helps to examine this problem by focusing again on Popper's collection of methodological rules: they purportedly help the scientific community direct themselves towards certain short-term ends, namely (1) eliminating error, and (2) increasing the empirical content of theoretical systems. These are two short-term strategies that are, in turn, aimed at long-term goals.

The first question is raised: why are these ends valuable? Are they valuable in-themselves? Valuable for non-epistemic ends? Valuable for epistemic (but not truth-related) ends? Or valuable for truth-related ends?

There is some tension in Popper's thought over the final aims of scientific practice: in The Logic of Scientific Discovery, Popper claims 'it is possible to avoid using the concepts "true" and "false"' in the construction of aims (2002, 273, cf. 275)--progress is viewed as a matter of finding theories with increasing explanatory and predictive power. This would eliminate (d) as the final value, leaving (a), (b) and (c) as candidate ends.

If one aligned themselves with 'early Popper', certain normative rules like aiming towards increasing the predictive and explanatory content of theories would be aimed at satisfying a 'supreme rule' (2002, 33) of making a certain set of speech-acts as empirically informative as possible. Furthermore, other rules insure this set of speech-acts are not protected against future testing in any way by refusing to adopt certain argumentative stratagems, such as making ad hoc adjustments to the theoretical system, all aimed at satisfying this 'supreme rule'.

The first question is raised once more: why is this 'supreme rule' valuable? Is it valuable in-itself? Valuable for non-epistemic ends? Valuable for epistemic (but not truth-related) ends?

The reasons to adopt the 'supreme rule' is nothing but 'a proposal for an agreement or convention' by individuals with certain argumentative or ethical standards, such as a desire to avoid dogmatism, a desire to speak informatively and address 'new and unexpected questions, challenging us to try out new and hitherto undreamed-of answers' (ibid. 15). This is an 'infinite yet attainable aim: that of ever discovering new, and more general problems, and of subjecting our ever tentative answers to ever renewed and ever more rigorous tests' (ibid. 281).

Note that there is no reference made to ultimate epistemic standards such as truth, but rather socially agreed-upon conventions that withstand criticism--including, but not exclusive to, the socially agreed-upon convention that a particular methodology is truth-conducive or truth-directed. In short, 'early Popper' grounds this 'supreme rule' in extra-epistemic grounds, namely (a) and (b). These simply are the 'rules of the game' of science. Popper presents these rules as candidates and offers them up for criticism. Other philosophers of science may improve on these rules, or show how these rules may be abandoned. By this process of conjecture and criticism, we can arrive at a (provisionally) stable position.

3. Why accept these 'rules of the game' for scientists?

On what grounds can any proposed 'rules of the game' for scientists be preferred? Popper's approach swallows all disciplines: if the 'rules of the game' for scientists are socially agreed-upon conventions that withstand criticism, what is good for the goose is good for the gander: philosophers of science subject aims to criticism in the hopes of improving or rejecting these aims, as well as proposing interesting and novel rules (and aims) that supplement or replace these rules and aims.

In short, Popper's approach is ecumenical: nothing at the onset forbids the introduction of a final value (such as approximating the truth), but furthermore nothing demands introducing a final value. It is simply up for philosophers of science (and scientists) to engage in critical disputation over which final value(s) should be preferred over others. Perhaps there are many, or merely one; perhaps it is agreed that it all bottoms out in convention, in ethical imperatives to treat others as rational agents deserving of a hearing, or there is the conventional agreement to take the final value as truth. It is up to philosophers to discuss amongst one another and--eventually--decide over what aims are desirable.

We have a second-order problem, namely, philosophers are faced with two intertwined paths: philosophers can (1) conventionally agree to accept or reject certain rules, and (2) conventionally agree to accept or reject certain aims. The moment, however, we take either rules or aims to be beyond question, we stop playing the 'game of philosophy'.

But why accept these 'rules of the game' for philosophers? Here is one answer: presumably, these 'rules of the game' are accepted for the same reason why the 'rules of the game' for scientists are adopted--these 'rules of the game' avoid dogmatism, increase the informativeness of our speech-acts and open up henceforth unarticulated and interesting questions. If there is something objectionable to any of these starting aims, they will be eventually uncovered, so long as philosophers of science and diligent, quick-witted and fortunate to discover some objection to these rules and aims.

Similarly, if there is something (henceforth unarticulated) valuable to be found in dogmatism, threats or violence, then they may be eventually uncovered, since these pathways are not closed from our starting point, either. However, these ways do not look promising, since they do not discriminate in their content: any position can be defended come what may by being dogmatic, threats or violence, and would (presumably) end the 'game of philosophy' prematurely by adopting them.

4. 'Later Popper'

By the time of the publication of Conjectures and Refutations (1963) 'later Popper' defends a different aim for scientific progress: an increase in explanatory and predictive power is valuable because it is aimed at truth (ibid. 226), for 'we simply cannot do without something like this idea of a better or worse approximation to truth' (ibid. 232). While 'early Popper' is a value-pluralist, and believed truth-as-correspondence may be one of many potential aims (but is not necessary from the onset), now 'later Popper' is a value-monist: the one final aim is, so Popper thought, approximation to truth.

It is clear that both approaches are compatible with Popper's later work on critical rationalism: without any strong fulcrum with which to lift up the rest of the intellectual scaffolding, there is no Wittgensteinian 'hinge proposition' or 'Archimedean point'; we are left with a radical form of fallibilism, the 'early' version on which begins with no reference to truth or falsity (although such an avenue is open to exploration) and the later version has gone down this route and concluded that the best available way to explain why one theoretical system is preferable to another is by recourse to their relative degrees of verisimilitude.

It is easy to see how adopting these methodological rules will reliably secure the ends of 'early Popper'; however, it is not as clear how 'late Popper' can. In fact, it seems an article of faith that any proposed methodological rule in the sciences will reliably aim towards the truth, without assuming some questionable metaphysical assumptions on the relationship between any features of a theoretical system--such as explanatory or predictive power--and truth. Value-monism is on shakier argumentative grounds than value-pluralism. That is, it remains an open question over whether the world is so constructed that broad and deep explanations are more likely to capture any underlying, hidden features of the world than narrow and shallow explanations.

If we are to be value-pluralists, and avoid dogmatism, desire to speak informatively and and always address new and interesting problems, without any final goal of truth, it is here that Feyerabend raises his strongest objection against normative demarcation criteria by (perhaps inadvertently) producing a reductio for the 'early Popper'.

5. Feyerabend's objection

Feyerabend introduced his aim of theoretical pluralism as 'an essential part of the empirical method' (1999a, 91): Feyerabend extends Popper's early work on critical experiments, noting that absent the existence of alternative theories, severe testing is likely to be impossible after some time, without good fortune. Without a rival theory, we are left in a holding pattern, without an idea of where to look for a severe test of a theoretical system. Feyereabend begins his argument for theoretical pluralism by attacking 'the assumption of the relative autonomy of facts, or the autonomy principle' (ibid, 91):

In Feyerabend's words, 'Not only is the description of every single fact dependent on some theory (which may, of course, be very different from the theory to be tested). There exist also facts which cannot be unearthed except with the help of alternatives to the theory to be tested, and which become unavailable as soon as such alternatives are excluded' (ibid. 91-21).

Thus '[t]he methodological unit to which we must refer when discussing questions of test and empirical content is constituted by a whole set of partly overlapping, factually adequate, but mutually inconsistent theories' (ibid. 92).

Consequently, we have a principle of proliferation, which aims at satisfying the ends set out by 'early Popper': 'Invent and elaborate theories which are inconsistent with the accepted point of view, even if the latter should happen to be highly confirmed and generally accepted' (1981, 105). In this way, scientists are rarely--if ever--caught in a holding pattern.

That is not to say that Feyerabend's theoretical pluralism on its own is at odds with Popper's later approach. In fact, Feyerabend credits Popper with being the first to point out 'the need, in the process of the refutation of a theory, for at least another theory' (1962, 32). Rather, it shows that Feyerabend extends both 'early Popper' and Popper's later approach by introducing his principle of proliferation: there are no prior restrictions laid down on what rival theories we can introduce, and we should do our best to produce as many as possible in order to develop surprising and novel ways to test them.

Other proposals by Feyerabend focus on the short-term, tactical level, such as, for example, his proposal called the 'principle of tenacity', dictating that scientists ought to stick to promising theories 'despite considerable difficulties' (1999b, 107).  This proposal seems to be an improvement to both 'early Popper' and 'later Popper', since it makes explicit Popper's observation that any crucial experiment is never uncontroversial: what appears to be a reductio may in fact be a surprising consequence of a theory. Think of, for example, how the Popper experiment and EPR experiment did not refute certain interpretations of quantum mechanics, but rather was a crucial experiment that corroborated these interpretations.

However, Feyerabend also included some eliminative proposals, such as in the first edition of Against Method, Feyerabend suggested removing the condition of 'factual adequacy' (1975, 41, fn. 8), thus repudiating later Popper's final aim of approximation to truth. Since traditional epistemic standards of aiming at truth are no longer paramount, all theories are 'scientific' in both a descriptive and normative sense, so long as they are empirically significant and add to the proliferation of theories.

Feyerabend's rejection of 'factual adequacy' can also be seen as an over-extension of Popper's critical rationalism to embrace 'a radically sceptical interpretation of Popper's own philosophy of science' (Lakatos 1978, 166, fn. 2): high-level theoretical systems are not just tested by appealing to other, accepted, low-level theoretical systems describing some space-time region K; they are now tested by appealing to other high-level theoretical systems. No longer are 'basic statements' limited to what is intersubjectively agreed-upon by recourse to experimentation and testing, but simply what is agreed-upon, irrespective to whether there is an appeal to an empirical basis of testing.

This proposal itself was criticised by Hilary Putnam as the claim that scientists ought to strive for 'alternative explanations by means of false theories', so that progress is not approximation to truth, but rather the explosion of a sea of theories which 'will later turn out to be false, and will be superseded by new batteries of false theories' (1975, 119).

It is here that Feyerabend's far more mature objection to Popper becomes clear: without the ability to appeal to any empirical evidence of the superiority of scientists maintaining truth as a regulative ideal or goal (for the REA, CEA and tactical objections are impermissible), on what grounds should we prefer Popper's goal of approximating truth over Feyerabend's far more practical goal of proliferation?

We may desire approximation of truth, but others may also desire nothing beyond proliferation. We are left with this choice, and there are no obvious arguments for one over the other; rather, it is a matter of aesthetic preference, rather than providing a rationale for preferring truth over proliferation. We are therefore caught in a bind: we can desire truth, but have no compelling arguments to believe any rules will help guide us towards truth; or we can desire proliferation, and aim at something easily obtainable.


Feyerabend, P.K. (1962). Explanation, reduction and empiricism. In H. Feigl, & G. Maxwell (Eds) Scientific explanation, space and time. Minneapolis: University of Minnesota Press: 28-97.

_______. (1975). Against Method. London: Verso.

_______. (1981.) Reply to criticism. Comments on Smart, Sellars and Putnam. In P.K. Feyerabend, Realism, rationalism and scientific method. Philosophical papers, Vol. 1. Cambridge: Cambridge University Press: 104--131.

_______. (1987). Farewell to Reason. London: Verso.

_______. (1999a). How to be a good empiricist. A plea for tolerance in matters epistemological. In P.K. Feyerabend, Knowledge, Science and Relativism. Philosophical Papers, Vol. 3. Ed. J. Preston. Cambridge: Cambridge University Press: 78--103.

_______. (1999b). Outline of a pluralistic theory of knowledge and action. In P.K. Feyabend, Knowledge, Science and Relativism.

Lakatos, I. (1978). Popper on demarcation and induction. In I. Lakatos, Philosophical Papers, Vol 1: The methodology of scientific research programmes. Ed J. Worrall and G. Currie. Cambridge: Cambridge University Press: 139-167.

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

______. (1963). Conjectures and Refutations. London: Routledge & Kegan Paul.

Putnam, H. (1975). How not to talk about meaning. In H. Putnam, Philosophical papers, Vol. 2: Mind, Language and Reality, Cambridge: Cambridge University Press. 117--131.