|A New Look at Falsification In Light Of the Duhem-Quine Thesis|
by Andrew Lewthwaite
As an answer to difficulties associated with the traditional inductive method, Karl Popper responds with falsification, a deductive scientific method where, as he describes it, "[a] scientist, whether theorist or experimenter, puts forward statements, or systems of statements, and tests them step by step." [i] In addition to serving as a solution to the problems of inductive methodology, falsification is also as a demarcation criterion between science and non-science. Popper's method relies heavily on hypothesis testing as the activity characteristic of scientific research. However, this activity would be challenged by the Duhem-Quine thesis. The consequences of this thesis inadvertently question falsification with respect to the feasibility of hypothesis testing, and as a result, challenges the method to the core, both as a means of expanding scientific knowledge and as a demarcation criterion. What will come to light in this paper is that falsification withstands most of the damage done by the Duhem-Quine thesis on the methodological level if certain modifications are made, but as a demarcation criterion, Popper's method cannot provide a clear cut distinction between scientific and non-scientific hypotheses.
The two difficulties that are addressed by Popper that led him to the development of his falsification method are the problems of induction and demarcation. This note is worthy of consideration as Popper's philosophy in its totality is often looked upon as focusing on problems, developing answers by way of creative theorizing, and then testing their validity in practical terms. In a similar manner, the easiest way to elucidate the 'why' and the 'what' of falsification is to examine the problems to which it is considered an answer, that it is considered an answer, and to look at the method in its evolution from those difficulties.
In addressing the first of the two problems, what is called into question is the traditional scientific method of induction. Following in the footsteps of David Hume [ii] and Bertrand Russell [iii] , Popper criticises induction because of its inability to justify itself on logical grounds. Without a principle of induction that would allow universal generalizations to be inferred from particular experiences, pure induction has no logical justification. However, even if such a principle were proposed, it would have to be substantiated by way of induction, with the existence of a further principle of the same nature that has the same problem. Faced with an infinitely regressing or a circular line of reasoning, the search for this form of justification of induction must be abandoned. One must make a decision as to the acceptability of the method. Either it is received on irrational grounds, or one chooses to be sceptical of its ability to provide justified scientific knowledge. Popper opted for the sceptic's route, and formulated his new methodology on deductive logic, which technically, can avoid the foundational disputes that plague the inductive method. Thus, for Popper, the scientist puts forth generalizations in the form of hypotheses, and then deduces conclusions in the form of predictions, which she can then test. It might be worth noting that Popper's falsification method concerns only the logical process of hypothesis testing, not where the hypothesis originates from. The creativity involved in theoretical scientific discovery is a psychological matter, one that escapes logic for the most part [iv] . Consequently, he claims that this is one of the biggest differences between himself and those who, in the past, favoured inductive methodology [v] , in which the distinction between the psychology of knowledge (i.e. the invention of theories) and the logic of knowledge (i.e. logical validity of the method) is not made.
Concerning the idea of testing a scientific hypothesis, Popper responds to the verificationist or confirmationist position propounded by the logical positivists such as Rudolph Carnap and A.J. Ayer [vi] . The verificationist position consists of the idea that a hypothesis gains truth content by passing tests in which it is confirmed by empirical evidence. In plain language, if observations agree with what was hypothesized, the scientist can declare that their hypothesis was verified by nature. For Popper, verifying theories is too simple, and he criticises science done in this manner for its unwillingness to challenge itself. Popper describes what he means here, by noting that "[i]t is easy to obtain confirmations, or verifications, for nearly every theory -- if we look for confirmations" [vii] and that, "[c]onfirmations should only count if they are the result of risky predictions; that is to say, if, unenlightened by the theory in question, we should have expected an event which was incompatible with the theory -- an event which would have refuted the theory." [viii] If a confirmation is to occur, it will only happen as a result of surviving a severe attempt at refutation.
Therefore, it is the notion of falsifiability or refutability that gives strength to scientific hypotheses. Through these tests, or attempted refutations, scientists check their conjectures about the nature of reality with what is really there. Furthermore, falsified hypotheses are encounters where the imaginative theorist learns that he was mistaken in what he thought to be the nature of reality, and hence he must proceed from his mistakes and form new conjectures about the scientific problems that are to be solved.
To turn to the other important aspect of Popper's theory, note once again that falsificationism is not only a methodology, but also a criterion for the distinction between science and non-science. Traditionally, the inductive method was what distinguished science from metaphysics in that a theory was required to make reference to empirical evidence that would verify its claims. In an attack on this idea, Popper points out that there are many theories that qualify as scientific (following the inductive criterion), but their explanatory power is too great; they are able to account for nearly any conceivable empirical event without difficulty. His disfavour of theories that escape legitimate objection lead him to the new formulation of the demarcation criterion. In his words, "[o]ne can sum up all this by saying that the criterion of the scientific status of a theory is its falsifiablility, or refutability, or testability." [ix]
There are two ways to look at this proclamation. First of all, in simple logical terms, Popper's strategy is more feasible. No amount of singular confirmations can ever prove the truth of a universal proposition because universals are infinite (they have no spatial or temporal limits) and particulars are finite (they have a specific spatial and temporal location). However, a single observation can bring down an entire generalization if it contradicts what that statement claims. [x] Using the ever so famous example, the observation of a single black swan, disproves the contention that "all swans are white". Now, the asymmetry Popper shows is very simple and can be avoided, but the main tenet of the argument is that falsification stands on stronger logical grounds.
Secondly, falsification requires that a theory leave itself open to refutation, any theory not doing so is considered unscientific or pseudo-scientific. This relates to theories that cannot be genuinely refuted by any conceivable event, and in Popper's words, "[appear] to be able to explain practically everything that [happens] within the fields to which they [refer]." [xi] This also relates to Popper's insistence that scientific knowledge is never final "truth", for finality concerning knowledge closes the door to refutation. As a result, theories such as Freud's psychoanalysis and Adler's individual psychology, which do not have the quality of refutability, are not scientific. [xii]
One of the more important results derived from Popper's new criterion is that all knowledge is to be considered tentative. If a hypothesis passes severe-testing, it is said to have been corroborated, temporarily. No theory can ever be proven true; there must always be the possibility of refutation, otherwise science will become dogmatic regarding a well-corroborated theory. As a result, a theory that purports to explain everything (and appears to stand up in practice) limits the progress of scientific research (i.e. nothing new is discovered or investigated, we simply add to the mass of confirmations for the theory). By adhering to the quality of refutability, this problem is avoided altogether.
Along these lines, it's not unreasonable to believe that if a theory is not continually challenged by imaginative and rigorous tests, observed phenomena (i.e. anomalies) that deviate from the dominant theory of the time might just simply be ignored or pushed aside. In line with that idea, a theory that is genuinely falsified, might be redefined in such a way that it escapes the situation in which it was shown to be problematic. [xiii]
To begin the analysis of falsification in light of the Duhem-Quine thesis, the nature of Popper's hypothesis testing will be briefly discussed, in order to set up the main criticisms that can be drawn out from Duhem and Quine's ideas. Popper's favourite example of attempted falsification concerns Eddington's eclipse observations in 1919 that brought about the first important confirmation of Einstein's relativity theory. Here he discusses this event:
Now the impressive thing about this case is the risk involved in a prediction of this kind. If observation shows that the predicted effect is definitely absent, the theory is simply refuted. The theory is incompatible with certain possible results of observation -- in fact with results which everybody before Einstein would have expected. [xiv]
This type of test provides a theory with either its own demise, or a solid piece of empirical evidence that corroborates its claims. In this case, Einstein postulated a hypothesis that predicted an event that could not be accounted for by Newton's theory. Einstein's theory went beyond the accepted Newtonian view, explained something new, provided itself with a solid piece of corroborative evidence and falsified an aspect of Newton's theory. The point here is that instead of simply supplying a theory with another check in the confirmation column, falsification aims at causing the theory to 'sweat out' its content by putting it through extremely challenging tests. [xv]
In a similar vein, Pierre Duhem, a French physicist and philosopher talked about 'crucial experiments', and it is on this note that the first criticism of Popper's methodology comes to light. A 'crucial experiment' is an experiment that establishes the truth of one among a set of competing theories. Now, one can see firsthand that Popper would object to the aims of such an experiment (i.e. establishing truth), but that does not affect its application to his theory. If two theories compete to explain the same thing, using the reductio ad absurdum principle, one theory is proven true by exposing the falsity, or contradiction, present in the other theory. [xvi] Now, from Popper's angle, one theory is corroborated by observations that falsify the other. This idea fits perfectly with the Einstein-Newton episode previously discussed.
However, there is a genuine problem. What the Duhem-Quine thesis clearly implies is that it is impossible to test a hypothesis in isolation. Rather, the entire system or group of hypotheses that make up the basis of a theory are subjected to experimental scrutiny. On top of this, a number of auxiliary assumptions, involving experimental equipment, conditions and substance are involved in the formulation of a test, and they are necessarily bound up in its results. Duhem explains the consequences of this idea in his own words:
In sum, the physicist can never subject an isolated hypothesis to experimental test, but only a whole group of hypotheses; when the experiment is in disagreement with his predictions, what he learns is that at least one of the hypotheses constituting this group is unacceptable and ought to be modified; but the experiment does not designate which one should be changed [xvii]
It is difficult to decide under what circumstances one would concede in rejecting a hypothesis (or an entire system of scientific statements) if it is impossible to pinpoint the location of the error. It is worth noting however, that, as in the case of the Eddington experiment, if an alternative theory or hypothesis can explain the deviation (as well as everything the 'falsified' theory explains), science tends to accept the newer theory while rejecting partially, if not totally, the other.
Now however, Popper is much too shrewd a philosopher to claim that the implications of this thesis do not affect his method of falsifiability. Thus he acknowledges the problem:
The attribution of the falsity to some particular hypothesis that belongs to this set of premises is therefore risky, especially if we consider the great number of assumptions which enter into every experiment......The answer is that we can indeed falsify only systems of theories and that any attribution of falsity to any particular statement within such a system is always highly uncertain [xviii]
Yet, the consequences of the Duhem-Quine thesis are not avoided by this assertion alone, although it is possible to emphasize the apparent requirement of an alternative explanation to legitimately falsify a theoretical system. It is unlikely that a scientist or theorist would accept the denouncement of an entire scientific theory simply on the grounds that an error has been identified in one of the many premises in an experiment test of that theory. Indeed, no legitimate 'knowledge hunter' would be so hasty in dismissing a life's amount of work.
Another aspect of the thesis, specifically addressed by Quine more so than Duhem concerns the modification of a theory that contains error. It refers to the idea of pinpointing where the error lies when a hypothesis clashes with the observation that was predicted from it's premises. Now, as has been stated through the Duhem-Quine thesis, because of the multitude of auxiliary hypotheses and assumptions, the hypothesis in question is never the 'sure-fire' source of the error, and cannot therefore be conclusively falsified. As a solution to this problem, the experiment could be altered in some way so as to exclude some auxiliary assumptions and welcome others to achieve the same prediction (minus the error). But this proves controversial for hypothesis testing. As Quine puts it, "[a] recalcitrant experience can, I have urged, be accommodated by any of various alternative re-evaluations in various alternative quarters of the total system;" [xix] . If the explanatory composition can be changed, and the same result obtained, then it appears that testing a hypothesis is a constructive process. Relying on principles of will or convention, certainly not logic, the scientist can arrange the hypothesis and auxiliary assumptions in such a way that the desired prediction is obtained.
To continue, if falsifiability is to be the criterion for what is scientific, then it must also answer to certain difficulties that result from the main conclusion of the Duhem-Quine thesis that hypotheses cannot be tested in isolation. The question is phrased as such: How do we judge the scientific status of a hypothesis (or law) if we cannot consider it apart from the system it belongs to? What is questioned is whether scientific hypotheses are falsifiable on their own, or whether they require the entire theoretical system from which they are derived, plus a set of auxiliary hypotheses in order to be put into testable form. Donald Gillies, in his book Philosophy of Science in the Twentieth Century, provides an example of this exact situation. He points out that Newton's first law cannot be tested in isolation, for it requires conjunction with Newton's second and third laws, the law of gravity, as well as numerous auxiliaryassumptions. [xx] Following Popper's demarcation criterion, Newton's first law would not be scientific since it is not falsifiable. Interestingly enough, in his criticisms of the verificationist criterion of demarcation, Popper claims that, "[t]his criterion excludes from the realm of meaning all scientific theories (or 'laws of nature');" [xxi] Unfortunately, as a result of the Duhem-Quine thesis, the laws of nature appear to suffer the same fate under Popper's criterion.
In response to this point, it is quite plausible to consider Newton's first law scientific when joined together with the rest of the theoretical system (i.e. with the second and third laws, the law of gravity and the auxiliaries). However, Gillies points out a problem associated with this idea: "[t]his modified demarcation criterion certainly allows Newton's first law to be scientific; but, unfortunately, it has the consequence that any arbitrary metaphysical statement is scientific." [xxii] The possible 'tacking' of metaphysical statements into theoretical systems would indeed become a problem. [xxiii] This conclusion is certainly inadequate, and leaves two options for consideration: 1) exclude some laws of nature from scientific status, or; 2) allow some theoretical statements to be considered scientific even though they are not falsifiable. [xxiv]
Unfortunately, the result of this dichotomy appears to indicate that falsifiability is too strict in its delineation of what is scientific and what is not. If falsification is to remain as the demarcation criterion, then one apparently must accept that some high level theories or hypotheses are unscientific. Be that as it may, there is another problem: assuming the scientist tests a theory as if it is scientific (in isolation), how does he or she decide what 'highly theoretical statements' are worthy of testing? With these ambiguous standards, it is arbitrary to choose Newton's first law over, for example, a statement relating to final causality in nature. Yet, it is highly unlikely that any scientist would leave Newton's first law off the table because it could not be classified as 'properly' scientific. Final causality, however, is not commonly regarded as 'scientific' in academic circles and is unlikely to warrant much interest, nevermind testing. It is simply the case that Popper's criterion is too narrow to accommodate statements of this nature (i.e. unfalsifiable in isolation).
The above objections may seem fatal to falsifiability as a demarcation criterion, but there is a different interpretation, which, does not 'deliver a fatal blow' to Popper's ideas. To do this, it is necessary to take a step back and examine, what seems to be a disposable distinction made by Gillies, regarding Popper's demarcation criterion. Concerning the idea of the testability of a hypothesis, Gillies claims certain universal laws are unscientific if falsification is regarded as the line of demarcation. However, it does not appear that Popper ever meant for specific hypotheses to fall under the criterion in isolation, but only as part of the 'system' they belong to.
Following along this line, Popper's criticism of the verificationist demarcation criterion that it excludes the laws of nature, is based on a criterion of meaning. In rejecting metaphysical statements as meaningless, the verificationists do the same to the laws of nature. Popper emphasizes this point:
But as I pointed out in my criticism the theory was based on a naïve or 'naturalistic' view of the problem of meaningfulness; moreover its propagators, in their anxiety to oust metaphysics, failed to notice that they were throwing all scientific theories on the same scrap-heap as the 'meaningless' metaphysical theories. [xxv]
Popper's formulation of the demarcation is not based on meaning, but rather on refutability, therefore one should not confuse the two as applying to the same nature of statements.
Therefore, in reference to Newton's first law (which according to Gillies is the shining example of the failure of Popper's criterion), there is contention with the idea that, standing alone, one can conclusively determine whether or not it qualifies as scientific. As part of Newton's theory as a whole, yes, this law is certainly scientific, because the theory itself is refutable [xxvi] . Popper illustrates this point here, "[a]ccording to this view, which I still uphold, a system (my emphasis) is to be considered as scientific only if it makes assertions which may clash with observations; and a system is, in fact, tested by attempts to produce such clashes, that is to say by attempts to refute it." [xxvii] There is no mention of hypotheses in isolation anywhere in Popper's discussion of his demarcation criterion because hypotheses do not warrant consideration apart from their systemic membership.
However, this idea does not constitute a solution quite yet. If the above answer stands, the scientist is still faced with the fact that Newton's first law, in isolation, is non-testable. What might be the case is that its scientific character remains 'undetermined' on its own. With Popper's emphasis on systems, it seems that this may have been what he meant from the beginning. This also fits nicely with the conclusions of the Duhem-Quine thesis, which avoids even considering hypotheses in isolation.
What must be dealt with next, is how a scientist determines whether a hypothesis or theoretical system can be rejected. What is proposed, is an interpretation of falsification that shies away from the simple idea that a theory can be rejected if one of its hypotheses clashes with an observation. The problem is that it appears to constitute what Popper calls the ad hoc definition of a theory. As was mentioned previously, Popper is not in favour of this idea, however, as a result of the Duhem-Quine thesis, it appears he is left with little choice. Therefore, in order to retain falsification as a method to any degree at all, this type of re-interpretation must be allowed.
Now, to return to the problem associated with the circumstances under which a scientist might reject a theoretical system as a result of its being falsified, it is necessary to turn to a distinction made by Imre Lakatos in his article "Falsification and the Methodology of Scientific Research Programs". He coins the terms naive falsificationist and sophisticated falsificationist, in order to separate what will be called a theoretical acceptance of falsification, and a practical acceptance of it. The naive falsificationist operates on the simple basis that if an observation statement contradicts the hypothesis that is being tested, then the hypothesis is rejected. Given the Duhem-Quine thesis, this procedure is not possible. Lakatos argues that the sophisticated falsificationist provides a much better delineation, outlined here:
The sophisticated falsificationist regards a scientific theory T as falsified if and only if another theory T' has been proposed with the following characteristics: (1) T' has excess empirical content over T: that is, it predicts novel facts, that is, facts improbable in the light of, or even forbidden, by T; (2) T' explains the previous success of T, that is, all the unrefuted content of T is contained (within the limits of observational error) in the content of T'; and (3) some of the excess content of T' is corroborated. [xxviii]
Basically, Lakatos is elaborating on the earlier suggestion that a theory will not normally be falsified unless there is another theory that can explain, along with the actual falsifying instance, everything that was contained in the old theory. If we use the quoted formulation in an example, Einstein's theory would be T' and Newton's theory would be T, and sophisticated falsificationism can be applied to the Eddington experiment. This appears to be a suitable example of what Popper envisioned as the application of falsification, the only difference being that the focus seems to shift away from the actual falsification to the 'corroboration' of a new theory. In this way, it is clear that one can accept falsification in light of the impossibility of testing hypotheses in isolation. For this is related to systems, and moreover, not simply on the rejection of systems, but on the acceptance of one of a set of competing systems.
On top of that, the above suggestion provides an escape from the dangers of modifying erroneous theories. As Quine pointed out, a hypothesis test might be rearranged in order to avoid the "recalcitrant experiences" that suggest falsity in the premises of an experiment. As mentioned, this practice is very much a matter of convention, and seriously compromises any objective quality for scientific knowledge. Sophisticated falsificationism avoids the problems of experimental rearrangement in this sense, for the reason that it requires the existence of an alternative explanation for the experimental error. Falling into the convention-driven type of arbitrary procedure is what poses the most significant challenge to falsification as a whole, and Lakatos' brand of testing manages to avoid it altogether. Arbitrariness, dogmatism and conventionalism are all elements that Popper intended to avoid with his original formulation of falsificationism, and fortunately enough, it seems that this continues to be the case, even after Lakatos' reinterpretation.
One last thing to be addressed is the issue concerning the direction of science if the falsification method is followed. Many critics have attacked Popper on the contention that his methodology is counter-productive to the progress of science. They find it hard to believe that scientists valiantly attempt to prove their own theories wrong. Consequently, this is a naïve way to look at things, and Popper's true message gets lost in light of it. In positing the falsification method, although refutability is paramount, rejecting conjecture after conjecture is not the intended result. Rather, by submitting theories to the most excruciating tests imaginable (those which would be most likely to falsify them), any corroborating evidence that comes out of such tests will be hard earned, and solid. By leaving the door open to refutation, scientists can narrow down the areas of a theory that are mysterious, foggy and possibly troublesome for the rest of the whole. Corroboration, although tentative, is really where theories gain their explanatory power and become valuable to science, but this can only happen after significant challenges to the theories have been made. Realistically, if a scientist is searching for simple verifying instances, it is highly unlikely that she will attempt anything that might prove detrimental to their research. With refutability as the goal, no shortcuts will be taken.
To sum up, in light of the problematic Duhem-Quine thesis, Popper's falsification still stands up as a method, and to some degree, as a demarcation criterion. Of course, it may not retain its original polar opposition in face of the verificationist position that Popper would have liked. As well, some of the aims of falsification have been reinterpreted in order to coincide with the idea of 'corroboration'. But as a whole, the main tenets still hold true. More importantly, the brand of sophisticated falsification put forth by Lakatos protects the scientific method from falling into the trap created by experimental redefinition, that of a brand of conventional or pragmatic science. If science is to remain as a source of objective knowledge, it must be protected from these types of interpretations, which are characterized by arbitrary decisions. Thus, if anything, falsification is a defender of tentative objective scientific knowledge.
[i] Popper. 1959. p. 27.
[ii] See "The Problem of Induction" by David Hume in Readings in the Philosophy of Science. Edited by T. Schick Jr. (Mountain View, Cal. U.S.A: Mayfield, 2000) pgs. 38-41.
[iii] See "On Induction" by Bertrand Russell in his book The Problems of Philosophy. (New York: Oxford University Press, 1997). Pgs 60-70.
[iv] Think black coffee, hot baths, meditation, etc.
[v] Some of Popper's contemporaries, such as Rudolph Carnap, did make the distinction between the invention of a theory and the justification of it, but remained faithful to the inductive method nonetheless.
[vi] See "The Elimination of Metaphysics" by A.J. Ayer in Readings in the Philosophy of Science. Edited by T. Schick Jr. (Mountain View, Cal. U.S.A: Mayfield, 2000) pgs. 6-8.
[vii] Popper. 1965. p. 36.
[viii] Ibid. p. 36.
[ix] Ibid. p. 37.
[x] Popper put forth a logical rule stating that falsity flows from singular statements to universal statements but not vice versa. It is upon this rule that his logical asymmetry is dependent.
[xi] Popper. 1965. p. 34.
[xii] See pg. 33-36 of Conjectures and Refutations. In this section, Popper dismisses Adler's individual psychology, and Freud's psychoanalysis as unscientific theories on the basis of their irrefutability. He claims that the nature of those theories makes it impossible to test their conclusions.
[xiii] See Conjectures and Refutations pg. 37 beginning, "The Marxist theory…" In this section, Popper describes how the followers of Marxism reinterpreted the theory and the evidence that had genuinely falsified some of its predictions in order to avoid disagreement. Popper refers to this as ad hoc adjustment and claims that it is detrimental to the scientific status of a theory.
[xiv] Popper. 1965. p. 36.
[xv] Popper describes this as a theory "proving its mettle." See Popper, Logic of Scientific Discovery, pg. 53 and Section #82: "The Positive Theory of Corroboration: How a Hypothesis may 'Prove its Mettle'.".
[xvi] Popper would never claim that a theory is proven "true", rather, by passing the experimental test unscathed, the theory is temporarily 'corroborated'.
[xvii] Duhem. 2000. p. 55.
[xviii] Popper. 1983, p.187.
[xix] Quine. 1965. p. 212.
[xx] Gillies. 1993. p. 212
[xxi] Popper. 1965. p. 261.
[xxii] Gillies. 1993. p. 213.
[xxiii] If, for instance, all statements in a theoretical system need not be falsifiable in isolation, what is to stop us from including statements such as "The Absolute is very hungry" from the group. We could simply add such a statement to the necessary components of the premises of an experiment, and derive the same conclusions to the experiment that would have been the case without the metaphysical statement. This is a problem, because a successful experimental outcome, would technically 'corroborate' or 'confirm' all the statements in the premises, including the metaphysical one that really had no role to play in the experiment.
[xxiv] Gillies proposes this solution, and adds his own interpretation of a demarcation criterion. (Gillies, pgs. 214-218)
[xxv] Popper. 1965. p. 259.
[xxvi] This point is admitted by Gillies himself on page 212 of Philosophy of Science in the Twentieth Century.
[xxvii] Popper. 1965. p. 256.
[xxviii] Lakatos. 2000. p. 21.
Ammerman, Robert (ed). Classics of Analytic Philosophy, (McGraw Hill: 1965)
Gillies, Donald. Philosophy of Science in the Twentieth Century. (Oxford:
Popper, Karl. The Logic of Scientific Discovery. (London: Hutchison, 1959)
Popper, Karl. Conjectures and Refutations. (New York: Harper and Row, 1965)
Popper, Karl. Realism and the Aim of Science. (Totowa, New Jersey: Rowman and
Russell, Bertrand. The Problems of Philosophy. (New York: Oxford University Press,
Schick Jr., Theodore (ed.) Readings in the Philosophy of Science. (Mountain View, Cal.
U.S.A: Mayfield, 2000)