170 Part 3 The Validation of Scientific Knowledge

Falsification and the

Methodology of Scientific
Research Programmes 1
Imre Lakatos

SCIENCE: REASON OR RELIGION? FALLIBILISM VERSUS FALSIFICATIONISM

For centuries knowledge meant proven knowl- Dogmatic (or Naturalistic) Falsificationism.
edge-proven either by the power of the intellect The Empirical Basis
or by the evidence of the senses. Wisdom and
. . . According to the "justificationists" scientific
intellectual integrity demanded that one must de-
knowledge consisted of proven propositions. Hav-
sist from unproven utterances and minimize, even
ing recognized that strictly logical deductions en-
in thought, the gap between speculation and es-
able us only to infer (transmit truth) but not to
tablished knowledge. The proving power of the
prove (establish truth), they disagreed about the
intellect or the senses was questioned by the scep-
nature of those propositions (axioms) whose truth
tics more than two thousand years ago; but they
can be proved by extralogical means. Classica/
were browbeaten into confusion by the glory of
intellectua/ists (or "rational ists" in the narrow
Newtonian physics. Einstein's results again turned
sense of the term) admitted very varied-and
the tables and now very few philosophers or scien-
powerful-sorts of extralogical "proofs" by reve-
tists still think that scientific knowledge is, or can
lation, intellectual intuition, experience. These,
be, proven knowledge. But few realize that with
with the help of logic, enabled them to prove
this the whole classical structure of intellectual
every sort of scientific proposition. Classical em-
values falls in ruins and has to be replaced: One
piricists accepted as axioms only a relatively small
cannot simply water down the ideal of proven
set of "factual propositions" which expressed the
truth-as some logical empiricists do-to the
"hard facts." Their truth-value was established by
ideal of "probable truth" 2 or-as some sociolo-
experience and they constituted the empirical ba-
gists of knowledge do-to "truth by [changing]
sis of science. ln order to prove scientific theories
consensus." 3 . . .
from nothing else but the narrow empirical basis,
they needed a logic much more powerful than the
deductive logic of the classical intellectualists:
1 This paper is a considerably improved version of my [ 1968b]
"inductive /agie." Ali justificationists, whether in-
and a crude version of my [1970]. Sorne parts of the former are
here reproduced without change with the permission of the
tellectualists or empiricists, agreed that a singular
Editor of the Proceedings of the Aristotelian Society. ln the prep- statement expressing a "hard fact" may disprove a
aration of the new version 1received much help from Tad Beck-
man, Colin Howson, Clive Kilmister, Larry Laudan, Eliot
universal theory; . . . but few of them thought
Leader, Alan Musgrave, Michael Sukale, John Watkins and John that a finite conjunction of factual propositions
Worrall.
might be sufficient to prove "inductively" a uni-
2 The main contemporary protagonist of the ideal of "probable
truth" is Rudolf Carnap. For the historical background and a versai theory. 4
criticism of this position, cf. Lakatos [1968a].
3 The main contemporary protagonists of the ideal of "truth by
consensus" are Polanyi and Kuhn. For the historical back-
4
ground and a criticism of this position, cf. Musgrave [ 1969a]. lndeed, even some of these few shifted, following Mill, the
Musgrave [1969b] and Lakatos [1970]. rather obviously insoluble problem of inductive proof lof uni-

From Sections 1, 2, and 3 of "Falsification and the Methodology of Scientific Research Programmes"
by Imre Lakatos, in Criticism and the Growth of Know!edge, Imre Lakatos and Alan Musgrave, Eds.
(1970), pp. 91-92, 94-125, 127-138, 189-195. Reprinted by permission of Cambridge University
Press.
 Falsification and the Methodology of Scientific Research Programmes 171

Justificationism, that is, the identification of sists in uttering on/y high/y probable theories; or
knowledge with proven knowledge, was the dom- even in merely specifying, for each scientific the-
inant tradition in rational thought throughout the ory, the evidence, and the probability of the theory
ages. Scepticism did not deny justificationism: lt in the light of this evidence.
only claimed that there was (and could be) no Of course, replacing proof by probability was
proven knowledge and therefore no knowledge a major retreat for justificationist thought. But
whatsoever. For the sceptics "knowledge" was even this retreat turned out to be insufficient. lt
nothing but animal belief. Thus justificationist was soon shown, mainly by Popper's persistent
scepticism ridiculed objective thought and efforts, that under very general conditions al 1theo-
opened the door to irrationalism, mysticism, su- ries have zero probability, whatever the evidence;
perstition. al/ theories are not on/y equally unprovable but
This situation explains the enormous effort in- also equally improbable. 6
vested by classical rationalists in trying to save the Many philosophers still argue that the failure
svnthetical a priori principles of intellectualism to obtain at least a probabilistic solution of the
and by classical empiricists in trying to save the problem of induction means that we "throw over
certainty of an empirical basis and the validity of almost everything that is regarded as knowledge
inductive inference. For ail of them scientific hon- by science and common sense." 7 lt is against this
esty demanded that one assert nothing that is un- background that one must appreciate the dramatic
proven. However, bath were defeated: Kantians change brought about by falsificationism in evalu-
hy non-Euclidean geometry and by non-Newto- ating theories and, in general, in the standards of
nian physics, and empiricists by the logical impos- intellectual honesty. Falsificationism was, in a
sibility of establishing an empirical basis (as Kant- sense, a new and considerable retreat for rational
rans pointed out, facts cannot prove propositions) thought. But since it was a retreat from utopian
and of establishing an inductive logic (no logic standards, it cleared away much hypocrisy and
can infallibly increase content). lt turned out that muddled thought, and thus, in fact, it represented
al/ theories are equal/y unprovable. an advance.
Philosophers were slow to recognize this, for First 1 shall discuss a most important brand of
obvious reasons: Classical justificationists feared falsificationism: dogmatic (or "naturalistic") 8 falsi-
that once they conceded that theoretical science is ficationism. Dogmatic falsificationism admits the
unprovable, they would have also to concede that fallibility of al/ scientific theories without qualifi-
rt is sophistry and illusion, a dishonest fraud. The cation, but it retains a sort of infallible empirical
philosophical importance of probabilism (or "neo- basis. lt is strictly empiricist without being induc-
1ustificationism") lies in the denial that such a con- tivist: lt denies that the certainty of the empirical
cession is necessary. basis can be transmitted to theories. Thus dog-
Probabilism was elaborated by a group of matic falsificationism is the weakest brand of justi-
Cambridge philosophers who thought that al- ficationism.
though scientific theories are equally unprovable, lt is extremely important ta stress that admitting
they have different degrees of probability (in the [fortified] empirica/ counterevidence as a final ar-
-;ense of the calcul us of probability) relative to the biter against a theory does not make one a dog-
available empirical evidence. 5 Scientific honesty matic falsificationist. Any Kantian or inductivist
then requires Jess than had been thought: lt con- will agree to such arbitration. But bath the Kantian

versai from particular propositions) to the slightly less obviously 6

For a detailed discussion, cf. my [1968a], especially pp. 353
insoluble problem of proving particular factual propositions
ff.
tram other particular factual propositions.
7
Russell [1943], p. 683. For a discussion of Russell's justifica-
·The founding fathers of probabilism were intellectualists;
tionism, cf. my 11962], especially pp. 167 fi.
Carnap's later efforts to build up an emprricist brand of probabi-
8 For the explanation of this term, cf. below, footnote 14.
lism lailed. Cf. my [1968a], p. 367 and also p. 361. footnote 2.
 172 Part 3 The Validation of Scientific Knowledge

and the inductivist, while bowing to a negative Weyl put it: "I wish to record my unbounded ad-
crucial experiment, will also specify conditions of miration for the work of the experimenter in his
how to establish, entrench one unrefuted theory struggle to wrest interpretable facts from an un-
more than another. Kantians held that Euclidean yielding Nature who knows so well how to meet
geometry and Newtonian mechanics were estab- our theories with a decisive No-or with an unau-
lished with certainty; inductivists held they had dible Yes." 12 Braithwaite gives a particularly lucid
probability 1. For the dogmatic falsificationist, exposition of dogmatic falsificationism. He raises
'-l however, empirical counterevidence is the one the problem of the objectivity of science: "To
\ and on/y arbiter which may judge a theory. what extent, then, should an established scientific
The hallmark of dogmatic falsificationism is deductive system be regarded as a free creation of
then the recognition that ail theories are equally the human mind, and to what extent should it be
conjectural. Science cannot prove any theory. But regarded as giving an objective account of the
although science cannot prove, it can disprove: lt facts of nature?" His answer is: "The form of a
"can perform with complete logical certainty [the statement of a scientific hypothesis, and its use to
act of] repudiation of what is false,"g that is, there express a general proposition, is a human device;
is an absolutely firm empirical basis of facts which what is due to Nature are the observable facts
can be used to disprove theories. Falsificationists which refute or fail to refute the scientific hypothe-
provide new-very modest-standards of scien- sis . . . [ln science] we hand over ta Nature the
tific honesty: They are willing to regard a proposi- task of deciding whether any of the contingent
tion as "scientific" not only if it is a proven factual lowest-level conclusions are false. This objective
proposition, but even if it is nothing more than a test of falsity it is which makes the deductive sys-
falsifiable one, that is, if there are factual proposi- tem, in whose construction we have very great
tions available at the time with which it may clash, freedom, a deductive system of scientific hypothe-
or, in other words, if it has potential falsifiers. 111 ses. Man proposes a system of hypotheses: Nature
1 Scientific honesty then consists of specifying, disposes of its truth or falsity. Man invents a scien-
i in advance, an experiment such that if the result tific system, and then discovers whether or not it
, contradicts the theory, the theory has ta be given accords with observed fact. " 1 3
i up. 11 The falsificationist demands that once a According ta the /agie of dogmatic falsifica-
proposition is disproved, there must be no prevari- tionism, science grows by repeated overthrow of
cation: The proposition must be unconditionally theories with the help of hard facts. For instance,
rejected. To (non-tautologous) unfalsifiable propo- according to this view, Descartes's vortex theory
sitions the dogmatic falsificationist gives short of gravity was refuted-and eliminated-by the
shrift: He brands them "metaphysical" and denies fact that planets moved in ellipses rather than in
them scientific standing. Cartesian circles; Newton's theory, however, ex-
Dogmatic falsificationists draw a sharp demar- plained successfully the then av ai !able facts, both
cation between the theoretician and the experi- those which had been explained by Descartes's
menter: The theoretician proposes, the experi- theory and those which refuted it. Therefore New-
menter-in the name of Nature-disposes. As ton's theory replaced Descartes's theory. Analo-

12
9 Quoted in Popper [19341. Section 85, with Popper's com-
Medawar [1967], p. 144.
ment: "I fully agree."
10
This discussion already indicates the vital importance of a 13
demarcation between provable factual and unprovable theoreti- Braithwaite [1953], pp. 367-8. For the "incorrigibility" of
cal propositions for the dogmatic falsificationist. Braithwaite's observed facts, cf. his [ 19381. While in the quoted
passage Braithwaite gives a forceful answer to the problem of
11
"Criteria of refutation have to be laid down beforehand: lt scientific ob1ectivity, in another passage he points out thdt "ex-
must be agreed which observable situations, if actually ob- cept for the straightforward generalizations of observable facts
served, mean that the theory is refuted" (Popper [1963], p. 38, . complete refutation is no more possible than 1s complete
foot note 3). proof" 1[19531. p. 19J.
 Falsification and the Methodology of Scientific Research Programmes 173

gously, as seen by falsificationists, Newton's the- But bath assumptions are false. Psychology
ory was, in turn, refuted-proved false-by the testifies against the first, logic against the secon(J,
anomalous perihelion of Mercury, while Einstein's and, finally, methodological judgment testifies
explained that too. Thus science proceeds by bold
speculations, which are never proved or even
made probable, but some of which are later elimi-
against the demarcation criterion. 1 shall discuss
them in turn.
-
nated by hard, conclusive refutations and then re- 1. A first glance at a few characteristic examples
placed by still bolder, new and, at least at the already undermines the first assumption. Gali-
start, unrefuted speculations. leo claimed that he could "observe" moun-
Dogmatic falsificationism, however, is unten- tains on the moon and spots on the sun and
able. lt rests on two false assumptions and on a tao that these "observations" refuted the time-ho-
narrow criterion of demarcation between scien- noured theory that celestial bodies are faultless
tific and non-scientific. crystal balls. But his "observations" were not
The first assumption is that there is a natural "observational" in the sense of being observed
psychological borderline between theoretical o~ by the-unaided-senses: Their reliability de-
speculative propositions on the one hand and fac- pended on the reliability of his telescope-
tuai or observational (or basic) propositions on the and of the optical theory of the telescope-
other. (1 shall call this-following Popper-the which was violently questioned by his
naturalistic doctrine of observation.) contemporaries. lt was not Galileo's-pure,
The second assumption is that if a proposition untheoretical-observations that confronted
satisfies the psychological criterion of being fac- Aristotelian theory but rather Galileo's "obser-
tuai or observational (or basic) then it is true; one vations" in the light of his optical theory that
may say that it was proved from facts. (1 shall call confronted the Aristotelians' "observations" in
the light of their theory of the heavens . . . .
this the doctrine of observationa/ (or experimenta/)
proof.) 14 This leaves us with two inconsistent theories,
prima facia on a par. Sorne empiricists may
These two assumptions secure for the dog-
concede this point and agree that Galileo's
matic falsificationist's deadly disproofs an empiri-
"observations" were not genuine observa-
cal basis from which proven falsehood can be car-
ried by deductive logic to the theory under test. tions; but they still hold that there is a "natural
These assumptions are complemented by a de- demarcation" between statements impressed
marcation criterion: Only those theories are "sci- on an empty and passive mind directly by the
entific" which forbid certain observable states of senses-only these constitute genuine "imme-
diate knowledge" -and between statements
affairs and therefore are factually disprovable. Or,
a theory is "scientific" is it has an empirical ba- which are suggested by impure, theory-im-
sis. 15 pregnated sensations. . . . But it transpires
from the work of Kant and Popper . . . that
. . . there are and can be no sensations unim- !
pregnated by expectations and therefore there i vf>.
!
14
For these assumptions and their criticism, cf. Popper [1934], 1
Sections 4 and 1O. lt is because of this assumption that-follow- is no natural (i.e., psychologica/) demarcation
1ng Popper-1 call this brand of falsificationism "naturalistic."
Popper's "basic propositions" should not be confused with the
between observational and theoretical propo-
basic propositions discussed in th1s section; cf. below, footnote sitions . . . .
35.
lt 1s important to point out that these two assumptions are 2. But even if there was such a natural demarca-
also shared by many justificationists who are not falsifica-
tionists: They may add to experimental proofs "intuitive
tion, logic would still destroy the second as-
proofs" -as did Kant-or "inductive proofs" -as did Mill. Our sumption of dogmatic falsificationism. For the
falsificationist accepts experimental proofs on/y.
truth-value of the "observational" proposi-
15 The empirical basis of a theory is the set of its potential falsifi-
tions cannot be indubitably decided: No fac-
ers: the set of those observational propositions which may dis-
prove 1t. tua/ proposition can ever be proved from an
174 Part 3 The Validation of Scientific Knowledge

experiment. Propositions can only be derived To support this last contention, 1 shall first tell
from other propositions, they cannot be de- a characteristic story and then propose a general
rived from facts: One cannot prove statements argument.
from experiences-"no more than by thump- The story is about an imaginary case of plane-
ing the table." 16 This is one of the basic points tary misbehaviour. A physicist of the pre-Einstein-
of elementary logic, but one which is under- ian era takes Newton's mechanics and his law of
stood by relatively few people even today. 17 gravitation (N), the accepted initial conditions, /,
If factual propositions are unprovable, then and calculates, with their help, the path of a newly
they are fallible. If they are fallible, then discovered small planet, p. But the planet deviates
clashes between theories and factual proposi- from the calculated path. Does our Newtonian
tions are not "falsifications" but merely incon- physicist consider that the deviation was forbid-
sistencies. Our imagination may play a greater den by Newton's theory and therefore that, once
raie in the formulation of "theories" than in established, it refutes the theory N? No. He sug-
the formulation of "factual propositions", 18 gests that there must be a hitherto unknown planet
but they are both fallible. Thus we cannot p' which perturbs the path of p. He calculates the
prove theories and we cannot disprove them mass, orbit, etc., of this hypothetical planet and
either. 19 The demarcation between the soft, then asks an experimental astronomer to test his
unproven "theories" and the hard, proven hypothesis. The planet p' is so small that even the
"empirical basis" is nonexistent: Ali proposi- biggest available telescopes cannot possibly ob-
tions of science are theoretical and, incurably, serve it: The experimental astronomer applies for a
fallible . . . . research grant to build yet a bigger one. 20 ln three
3. Finally, even if there were a natural demarca- years' time the new telescope is ready. Were the
tion between observation statements and theo- unknown planet p' to be discovered, it would be
ries, and even if the truth-value of observation hailed as a new victory of Newtonian science. But
statements could be indubitably established, it is not. Does our scientist abandon Newton's the-
dogmatic falsificationism would still be useless ory and his idea of the perturbing planet? No. He
for eliminating the most important class of suggests that a cloud of cosmic dust hides the
what are commonly regarded as scientific the- planet from us. He calculates the location and
ories. For even if experiments could prove ex- properties of this cloud and asks for a research
perimental reports, their disproving power grant to send up a satellite to test his calculations.
would still be miserably restricted: Exact/y the Were the satellite's instruments (possibly new
most admired scientific theories simply fail to ones, based on a little-tested theory) to record the
forbid any observable state of affairs. existence of the conjectural cloud, the result
would be hailed as an outstanding victory for
Newtonian science. But the cloud is not found.
Does our scientist abandon Newton's theory, to-
16 Cf. Popper [19341, Section 29. gether with the idea of the perturbing planet and
17 lt seems that the first philosopher ta emphasize this might the idea of the cloud which hides it? No. He sug-
have been Fries in 1837 (cf. Popper [1934], Section 29, foot-
note 3). This is of course a special case of the general thesis that
logical relations, like probability or consistency, refer ta propo-
sitions. Thus, for instance, the proposition "nature is consis-
tent" is false (or, if you wish, meaningless), for nature is nota
20
proposition (or a conjunction of propositions). If the tin y conjectural pl anet were out of the reach even of the
18 biggest possible optical telescopes, he might try some quite
lncidentally, even this is questionable. Cf. below, pp. 188ff.
navel instrument (like a radiotelescope) in order ta enable him
19 As Popper put it, "No conclusive disproof of a theory can
ta "observe it," that is, ta ask Nature about it, even if only
ever be produced"; those who wait for an infallible disproof indirectly. (The new "observational" theory may itself not be
before eliminating a theory will have ta wait forever and "will properly articulated, let alone severely tested, but he would
never benefit from experience" ([1934], Section 9). care no more than Galileo did.)
 Falsification and the Methodology of Scientific Research Programmes 175

gests that there is some magnetic field in that re- is inconsequential for the specific theory under
gion of the universe which disturbed the instru- test because by replacing the ceteris paribus
ments of the satellite. A new satellite is sent up. clause by a different one the specific theory can
Were the magnetic field to be found, Newtonians always be retained whatever the tests say.
would celebrate a sensational victory. But it is not. If so, the "inexorable" disproof procedure of
ls this regarded as a refutation of Newtonian sci- dogmatic falsificationism breaks down in these
ence? No. Either yet another ingenious auxiliary cases even if there were a firmly established em-
hypothesis is proposed or . . . the whole story is pirical basis to serve as a launching pad for the
buried in the dusty volumes of periodicals and the arrow of the modus ta/Jens: The prime target re-
story never mentioned again. 21 mains hopelessly elusive. 24 And as it happens, it is
This story strongly suggests that even a most exactly the most important, "mature" theories in
respected scientific theory, like Newton's dy- the history of science which are prima facie undis-
namics and theory of gravitation, may fail to forbid provable in this way. 25 Moreover, by the standards
any observable state of affairs. 22 lndeed, some sci- of dogmatic falsificationism ail probabilistic theo-
entific theories forbid an event occurring in some ries also corne under this head, for no finite sam-
specified finite spatio-tempora/ region (or briefly, a ple can ever disprove a universal probabilistic the-
"singular event") on/y on the condition that no ory26; probabilistic theories, like theories with a
other factor (possibly hidden in some distant and ceteris paribus clause, have no empirical basis.
unspecified spatio-temporal corner of the uni- But then the dogmatic falsificationist relegates the
versel has any influence on it. But then such theo- most important scientific theories on his own ad-
ries never a/one contradict a "basic" statement: mission to metaphysics where rational discus-
They contradict at most a conjunction of a basic sion-consisting, by his standards, of proofs and
statement describing a spatio-temporally singular disproofs-has no place, since a metaphysical
event and of a universal non-existence statement theory is neither provable nor disprovable. The
saying that no other relevant cause is at work any- demarcation criterion of dogmatic falsificationism
where in the universe. And the dogmatic falsifica- is th us still strongly antitheoretical.
tionist cannot possibly claim that such universal (Moreover, one can easily argue that ceteris
non-existence statements belong to the empirical paribus clauses are not exceptions, but the rule in
basis-that they can be observed and proved by science. Science, after ail, must be demarcated
experience. from a curiosity shop where funny local-or cos-
Another way of putting this is to say that some mic-oddities are collected and displayed. The
scientific theories are normally interpreted as con- assertion that "ail Britons died from Jung cancer
taining a ceteris paribus clause 23 : in such cases it between 1950 and 1960" is logically possible and
is always a specific theory together with this might even have been true. But if it has been only
clause which may be refuted. But such a refutation an occurrence of an event with minute probabil-
ity, it would have only curiosity value for the
crankish fact-collector; it would have a macabre
21
At least not until a new research program supersedes New-
ton's programme which happens to explain this previously re-
calcitrant phenomenon. ln this case, the phenomenon will be
24
unearthed and enthroned as a "crucial experiment". lncidentally, we might persuade the dogmatic falsificationist
22 that his demarcation criterion was a very naive mistake. If he
Popper asks, "What kind of clinical responses would refute
gives it up but retains his two basic assumptions, he will have to
to the satisfaction of the analyst not merely a particular diagno-
ban theories from science and regard the growth of science as
sis but psychoanalysis itself?" ([19631, p. 38, footnote 3.) But
an accumulation of proven basic statements. This indeed is the
what kind of observation would refute to the satisfaction of the
final stage of classical empiricism alter the evaporation of the
Newtonian not merely a part1cular version but Newtonian the-
hope that facts can prove or at least disprove theories.
ory itself?
25 This is no coincidence. . ..
23 This ceteris paribus clause must not normally be interpreted

as a separate premise. . 26 Cf. Popper [1934], Chapter VIII.

176 Part 3 The Validation of Scientific Knowledge

entertainment value, but no scientific value. A gress is pseudo-progress, that most, if not ail, of
proposition might be said to be scientific only if it the work do ne is irrational. If, however, sti Il ac-
ai ms at expressing a causal connection: Such con- cepting the demarcation criterion of dogmatic
nection between being a Briton and dying of Jung falsificationism, we deny that facts can prove
cancer may not even be intended. Similarly, "Ali propositions, then we certainly end up in com-
swans are white," if true, would be a mere curios- plete scepticism: Then ail science is undoubtedly
ity unless it asserted that swanness causes white- irrational metaphysics and should be rejected. Sci-
ness. But then a black swan would not refute this entific theories are not on/y equally unprovable,
proposition, since it may only indicate other and equa/ly improbable, but they are a/sa Pqual/y
causes operating simultaneously. Thus "ail swans undisprovab/e. But the recognition that not onlv
are white" is either an oddity and easily disprov- the theoretical but al/ the propositions in ~cience
able or a scientific proposition with a ceteris pari- are fal 1ible means the total col lapse of al/ torms oi
bus clause and therefore undisprovable. Tenacity dogmatic falsificationism as theories of sc1entiiic
of a theory against empirical evidence would then rational ity.
be an argument for rather than against regarding it
as "scientific." "lrrefutability" would become a
hal/mark of science.) . . .
Methodological Falsificationism. The
To sum up: Classical justificationists only ad-
"Empirical Basis."
mitted proven theories; neoclassical justifica-
tionists probable ones; dogmatic falsificationists The collapse of dogmatic falsificationism because
realized that in either case no theories are admissi- of fallibilistic arguments seems to bring us back to
ble. They decided to admit theories if they are square one. If al/ scientific statements are fallible
disprovable-disprovable by a finite number of theories, one can criticize them only for inconsis-
observations. But even if there were such disprov- tency. But then, in what sense, if any, is <,cience
able theories-those which can be contradicted empirical? If scientific theories are neither prov-
by a finite number of observable facts-they are able, nor probabilifiable, nor disprovable, then
still logically too near to the empirical basis. For the sceptics seem to be finally right: Science is no
instance, on the terms of the dogmatic falsifica- more than vain speculation and there is no such
tionist, a theory like "Ali planets move in ellipses" thing as progress in scientific knowledge. Can we
may be disproved by five observations; therefore still oppose scepticism? Can we save sc1entific
the dogmatic falsificationist will regard it as scien- criticism from fal/ibilism? ls it possible to have a
tific. A theory like "Ali planets move in circles" fallibilistic theory of scientific progress? ln particu-
may be disproved by four observations; therefore lar, if scientific criticism is fallible, on what
the dogmatic falsificationist will regard it as still ground can we ever eliminate a theory?
more scientific. The acme of scientificness will be A most intriguing answer is provided by meth-
a theory like "Ali swans are white," which is dis- odologica/ falsificationism. Methodological falsifi-
provable by one single observation. On the other cationism is a brand of conventionalism; there-
hand, he will reject ail probabilistic theories to- fore, in order to understand it, we must first
gether with Newton's, Maxwell's, Einstein's theo- discuss conventionalism in general.
ries, as unscientific, for no finite number of obser- There is an important demarcation between
vations can ever disprove them. "passivist" and "activist" theories of knowledge.
If we accept the demarcation criterion of dog- "Passivists" hold that true knowledge is l\Jc1ture\
matic falsificationism, and also the idea that facts imprint on a perfectly inert mind: Mental activlt)
can prove "factual" propositions, we have to de- can only result in bias and distortion. The most
clare that the most important, if not ail, theories influential passivist school is classical empir1cisrn.
ever proposed in the history of science are meta- "Activists" hold that we cannot read the book or
physical, that most, if not ail, of the accepted pro- Nature without mental activity. without interpret-
 Falsification and the Methodology of Scientific Research Programmes 177

ing 1t in the light of our expectations or theo- tablished theories: As science grows, the power of
ries.27 Now conservative "activists" hold that we empirica/ evidence diminishes. 29
are born with our basic expectations; with them Poincaré's critics refused to accept his idea,
we turn the world into "our world" but must then that, although the scientists build their conceptual
live for ever in the prison of our world. The idea frameworks, there cornes a time when these
that we live and die in the prison of our "concep- frameworks turn into prisons which cannot be de-
tual frameworks" was developed primarily by molished. This criticism gave rise to two rival
Kant; pessimistic Kantians thought that the real schools of revolutionary conventiona/ism: Du-
world is forever unknowable because of this hem's simplicism and Popper's methodological
prison, while optimistic Kantians thought that Cod falsificationism. 30
created our conceptual framework to fit the Ouhem accepts the conventionalists' position
world . . . . But revolutionary activists believe that no physical theory ever crumbles merely un-
that conceptual frameworks can be developed and der the weight of "refutations," but daims that it
also replaced by new, better ones; it is we who still may crumble under the weight of "continuai
create our "prisons" and we can also, critically, repairs, and many tangled-up stays" when "the
demolish them . . . . worm-eaten columns" cannot support "the totter-
New steps from conservative to revolutionary ing building" any longer31 ; then the theory loses
activism were made by Whewell and then by its original simplicity and has to be replaced. But
Poincaré, Milhaud and Le Roy . . . . Poincaré, falsification is then left to subjective taste or, at
Milhaud and Le Roy . . . preferred to explain the best, to scientific fashion, and leaves too much
continuing historical success of Newtonian me- leeway for dogmatic adherence to a favorite the-
chanics by a methodological decision taken by ory . . . .
scientists: After a considerable period of initial Popper set out to find a criterion which is both
empirical success scientists may decide not to al- more objective and more hard-hitting. He could
low the theory to be refuted. Once they have not accept the emasculation of empiricism, inher-
taken this decision, they solve (or dissolve) the ent even in Duhem's approach, and proposed a
apparent anomalies by au xi 1iary hypotheses or methodology which allows experiments to be
other "conventionalist stratagems." 28 This con- powerful even in "mature" science. Popper's
servative conventionalism has, however, the dis-
advantage of making us unable to get out of our
self-imposed prisons, once the first period of trial-
29 Poincaré first elaborated his conventionalism only with re-
and-error is over and the great decision taken. lt gard to geometry (ci. his [1891 ]). Then Milhaud and Le Roy
cannot solve the problem of the elimination of generalized Poincaré's idea to caver ail branches of accepted
physical theory. Poincaré's [1902] starts with a strong criticism
those theories which have been triumphant for a of the Bergsonian Le Roy against whom he defends the empiri-
long period. According to conservative conven- cal (falsifiable or "inductive") character of ail physics except for
geometry and mechanics. Duhem, in turn, criticized Poincaré:
tionalism, experiments may have sufficient power ln his view there was a possibility of overthrowing even Newto-
to refute young theories, but not to refute old, es- nian mechanics.
30
The loci classici are Duhem's [19051 and Popper's [19341.
Duhem was not a consistent revolutionary conventionalist.
Very much like Whewell, he thought that conceptual changes
are only preliminaries to the final-if perhaps distant-"natural
27
This demarcation-and terminology-is due to Popper; cf. classification":"The more a theory is perfected, the more we
especially his [1934), Section 19 dnd his [19451, Chapter 23 apprehend that the logical order in which it arranges experi-
and footnote 3 to Chapter 25. mental laws is the reilection of an ontological order." ln partic-
28 ular, he refused to see Newton's mechanics actually "crum-
Cf. especially Poincaré [18911 and [1902]; Milhaud [18961;
bling" and characterized Einstein's relativity theory as the
Le Roy [1899) and [1901 J. lt was one of the ch1ef philosophicdl
manifestation of a "frantic and hectic race in pursuit of a navel
merits of conventionalists to direct the limelight to the fact that
idea" which "has turned physics into a real chaos where logic
any theory can be saved by "conventionalist stratagems" from
loses its way and commonsense runs away frightened" (Pref-
refutations. (The term conventionalist stratagem is Popper\; d.
ace-of 1914-to the second edition of his [1905]).
the critical discussion of Poincaré's conventionalism in his
31
[1934), especially Sections 19 and 20. 1 Duhem [1905), Chapter VI, Section 10.
178 Part 3 The Validation of Scientific Knowledge

methodological falsificationism is both conven- that in the "experimental techniques" of the scien-
tionalist and falsificationist, but he "differs from tist fallible theories are involved, 36 "in the light of
the [conservative) conventionalists in holding that which" he interprets the facts. ln spite of this he
the statements decided by agreement are not [spa- "applies" these theories, he regards them in the
tio-temporally] universal but [spatio-temporally] given context not as theories under test but as un-
singular" 32 ; and he differs from the dogmatic falsi- problematic background knowledge, "which we
ficationist in holding that the truth-value of such accept (tentatively) as unproblematic while we are
statements cannot be proved by facts but, in some testing the theory." 3- He may call these theories-
cases, may be decided by agreement. 33 and the statements whose truth-value he decides
The conservative conventionalist (or method- in their light-"observational," but this is only a
ological justificationist, if you wish) makes unfalsi- manner of speech which he inherited from natu-
fiable by fiat some (spatio-temporally) universal ralistic falsificationism. 38 The methodological
theories, which are distinguished by their explana- falsificationist uses our most successful theories as
tory power, simplicity, or beauty. Our revolution- extensions of our senses and widens the range of
ary conventionalist (or "methodological falsifica- theories which can be applied in testing far be-
tionist") makes unfalsifiable by fiat some yond the dogmatic falsificationist's range of strictly
(spatio-temporally) singular statements which are observational theories. For instance, let us imag-
distinguishable by the fact that there exists at the ine that a big radio-star is discovered with a system
time a "relevant technique" such that "anyone of radio-star satellites orbiting it. We should like to
who has learned it" will be able to decide that the test some gravitational theory on this planetary
statement is "acceptable." 34 Such a statement may system-a matter of considerable interest. Now
be cal led an "observational" or "basic" state- let us imagine that Jodrell Bank succeeds in pro-
ment, but only in inverted commas. 35 lndeed, the viding a set of space-time co-ordinates of the plan-
very selection of al 1 su ch statements is a malter of ets which is inconsistent with the theory. We shall
a decision, which is not based on exclusively psy- take these statements as potential falsifiers. Of
chological considerations. This decision is then course, these basic statements are not "observa-
followed by a second kind of decision concerning tional" in the usual sense but only " 'observa-
the separation of the set of accepted basic state- tional .' " They describe pl a nets that neither the
ments from the rest. human eye nor optical instruments can reach.
These two decisions correspond to the two as- Their truth-value is arrived at by an "experimental
sumptions of dogmatic falsificationism. But there technique." This "experimental technique" is
are important differences. First, the methodologi- based on the "application" of a well-corroborated
cal falsificationist is nota justificationist, he has no theory of radio-optics. Cal 1ing these statements
illusions about "experimental proofs" and is fully "observational" is no more than a manner of say-
aware of the fal libi 1ity of his decisions and the risks ing that, in the context of his problem, that is, in
he is taking. testing our gravitational theory, the methodologi-
The methodological falsificationist realizes cal falsificationist uses radio-optics uncritically, as
"background knowledge." The need for decisions
ta demarcate the theory under test from unprob-
lematic background knowledge is a characteristic
32 Popper [1934], Section 30.
33 ln this section / discuss the "naive" variant ai Popper's meth-
odo/ogica/ falsificationism. Thus, throughout the section ··meth-
odologica/ ialsificationism" stands for "naive methodo/ogical 16 Cf. Popper 11934], end of Section 26 and also his [ 1 %8], pp.
talsificat1on1sm"; for this "naivety," cf. be/ow, pp. 181- 182.
291-2.
14 Popper [1934], Section 27. 37 Cf. Popper [1963], p. 390.
35Op cit. Section 28. For the non-basicness of these methodo-
18 lndeed, Popper carefully puts "observational" in quotes; cf.
logically "basic" statements, cf. e.g. Popper [1934] passim and
Popper [1959], p. 35, footnote *2. his [1934], Section 28.
 Falsification and the Methodology of Scientific Research Programmes 179

feature of this brand of methodologica/ falsifica- it may still be true. If we follow up this sort of
tionism. 39 (This situation does not really differ from "falsification" by the actual "elimination" of a
Gal ileo's "observation" of Jupiter's sa tel 1ites; theory, we may well end up by eliminating a true,
moreover, as some of Galileo's contemporaries and accepting a false, theory (a possibility which
rightly pointed out, he relied on a virtually non- is thoroughly abhorrent to the old-fashioned justifi-
existent optical theory-which then was less cor- cationist). . . .
roborated, and even less articulated, than present- The methodo/ogica/ fa/sificationist separates
day radio-optics. On the other hand, calling the rejection and disproof, which the dogmatic falsifi-
reports of our human eye "observational" only cationist had conflated . . . . He is a fallibilist but
indicates that we "rely" on some vague physiolog- his fallibilism does not weaken his critical stance:
ical theory of human vision. 40 ) He turns fallible propositions into a "basis" for a
This consideration shows the conventional el- hard-line policy. On these grounds he proposes a
ement in granting-in a given context-the new demarcation criterion: Only those theories-
(methodologically) "observational" status to a the- that is, non-"observational" propositions-that
ory.41 Similarly, there is a considerable conven- forbid certain "observable" states of affairs, and
tional element in the decision concerning the ac- therefore may be "falsified" and rejected, are "sci-
tuai truth-value of a basic statement which we take entific"; or, briefly, a theory is "scientific" (or
after we have decided which "observational the- "acceptable") if it has an "empirica/ basis." This
ory" to apply. One single observation may be the criterion brings out sharply the difference be-
stray result of some trivial error: ln order to reduce tween dogmatic and methodological falsifica-
such risks, methodological falsificationists pre- tionism. 44 . . .
scribe some safety contrai. The simplest such con- But even these three decisions are not suffi-
trai is to repeat the experiment (it is a matter of cient to enable us to "falsify" a theory which can-
convention how many times) . . . . not explain anything "observable" without a ce-
This is how the methodological falsificationist teris pari bus clause. 45 No finite number of
establishes his "empirical basis." (He uses quota- "observations" is enough to "falsify" such a the-
tion marks in order "to give ironical emphasis" to ory. However, if this is the case, how can one
the term.42) This "basis" can be hardly called a reasonably defend a methodology which daims to
"basis" by justificationist standards: There is noth- "interpret natural laws or theories as . . . state-
ing proven about it-it denotes "piles driven into ments which are partially decidable, i.e., which
a swamp." 43 lndeed, if this "empirical basis" are, for logical reasons, not verifiable but, in an
clashes with a theory, the theory may be called asymmetrical way, falsifiable . . . ?46 How can
"falsified", but it is not falsified in the sense that it we interpret theories like Newton's theory of dy-
is disproved. Methodological "falsification" is namics and gravitation as "one-sidedly decid-
very different from dogmatic falsification. If a the- able"?47 How can we make in such cases genuine
ory is falsified, it is proven false; if it is "falsified", "attempts to weed out false theories-to find the
weak points of a theory in order to reject it if it is
falsified by the test"? 48 How can we draw them

·Y This demarcation plays a raie bath in the first and in the

courth type of decisions of the methodological falsificationist

For the fourth decision, cf. below, p. 180.)
44
'
0
For a fascinating discussion, cf. Feyerabend [ 1969]. The demarcation criterion of the dogmatic falsificationist
was: A theory is "scientific" if it has an empirical basis (see
" One wonders whether it would not be better to make a break above, p. 173).
w1th the terminology of naturalistic falsificationism and rechris- 45
ten observational theories "touchstone theories." Cf. above, pp. 175-176.
46
2 Popper [1933].
' Popper [1963], p. 387.
47
Popper [1933].
" Popper [1934], Section 30; also cf. Section 29: "The Relativ-
48
•tv of Basic Statements." Popper [1957], p. 133.
180 Part 3 The Validation of Scientific Knowledge

into the realm of rational discussion? The method- consists in specifying, in advance, an experiment
ological falsificationist salves the problem by mak- such that if the result contradicts the theory, the
ing a further (fourth type) decision: When he tests theory has to be given up. 51
a theory together with a ceteris paribus clause and Methodological falsificationism represents a
finds that this conjunction has been refuted, he considerable advance beyond both dogmatic falsi-
must decide whether to take the refutation also as ficationism and conservative conventionalism. lt
a refutation of the specific theory. For instance, he recommends risky decisions. But the risb are dar-
may accept Mercury's "anomalous" perihelion as ing to the point of recklessness and one wonders
a refutation of the treble conjunction N 3 of New- whether there is no way of lessening them.
ton's theory, the known initial conditions, and the Let us first have a closer look at the risks in-
ceteris paribus clause. Then he tests the initial volved.
conditions "severely 49 and may decide to relegate Decisions play a crucial role in this methodol-
them into the "unproblematic background knowl- ogy-as in any brand of conventionalism. Deci-
edge." This decision implies the refutation of the sions, however, may lead us disastrously astray.
double conjunction N 2 of Newton's theory and The methodological falsificationist is the first to
the ceteris paribus clause. Now he has to take the admit this. But this, he argues, is the price which
crucial decision: whether to relegate also the ce- we have to pay for the possibility of progress.
teris paribus clause into the pool of "unproblem- One has to appreciate the dare-devil attitude
atic background knowledge." He will do so if he of our methodological falsificationist. He feels
finds the ceteris paribus clause well corroborated. himself to be a hero who, faced with two cata-
How can one test a ceteris paribus clause se- strophic alternatives, dared to reflect coolly on
verely? By assuming that there are other influenc- their relative merits and choose the lesser evi 1.
ing factors, by specifying such factors, and by test- One of the alternatives was sceptical fallibilism.
ing these specific assumptions. If many of them with its "anything goes" attitude, the despairing
are refuted, the ceteris paribus clause will be re- abandonment of ail intellectual standards, and
garded as wel 1 corroborated. . . . hence of the idea of scientific progress. Nothing
Thus, with the help of this fourth type of deci- can be established, nothing can be rejected, noth-
sion, so our methodological falsificationist has fi- ing even communicated: The growth of science 1-,
nally succeeded in interpreting even theories like a growth of chaos, a veritable Babel. For two thou-
Newton's theory as "scientific." . . . sand years, scientists and scientifically mindecl
To sum up: The methodological falsificationist phi losophers chose j ustificationist il lu~ions 01
offers an interesting solution to the problem of some kind to escape this nightmare . . . . Our
combining hard-hitting criticism with fallibilism. methodological falsificationist proudly reject'
Not only does he offer a philosophical basis for such escapism: he dares to measure up to the fui
falsification after fallibilism had pulled the carpet impact of fallibilism and yet escape skepticism b'
from under the feet of the dogmatic falsificationist, a daring and risky conventionalist policy, with no
but he also widens the range of such criticism very dogmas. He is fully aware of the risks but insisb
considerably. By putting falsification in a new set- that one has ta choose between some sort ot meth-
ting, he saves the attractive code of honour of the odologica/ falsificationism and irrationalism. Ht-
dogmatic falsificationist: that scientific honesty offers agame in which one has little hope of w1n-
ning, but daims that it is still better to plav than te
give up . . . .
49 For a discussion of this important concept of Popperian meth- But is not the firm strategy of the brand o·
odology, cf. my [1968aL pp. 397 ff. methodological falsificationism hitherto discussec:
so This type of decision belongs, in an important sense, to the
same category as the first decision: lt demarcates, by decision,
problematic from unproblematic knowledge. Cf. above p.
179, text to footnote 39. " See above, p. 172
 Falsification and the Methodology of Scientific Research Programmes 181

too firm? Are not the decisions it advocates bound it occurs in actual history is prima facie irrational
1
0 be tao arbitrary? Sorne may even claim that ail by the standards of our falsificationist. By his stan-
'.hat distinguishes methodological from dogmatic dards, scientists frequently seem to be irrationally
·alsificationism is that it pays lip-service ta fallibil- slow: For instance, eighty-five years elapsed be-
'm' tween the acceptance of the perihelion of Mercury
To criticize a theory of criticism is usual ly very as an anomaly and its acceptance as a falsification
::Jifficult. Naturalistic falsificationism was rela- of Newton's theory, in spite of the fact that the
'.1vely easy to refute, since it rested on an empirical ceteris paribus clause was reasonably well corrob-
~Jsychology of perception: One could show that it orated. On the other hand, scientists frequently
.\as simply fa/se. But how can methodological seem to be irrationally rash: For instance, Galileo
·alsificationism be falsified? No disaster can ever and his disciples accepted Copernican heliocen-
disprove a non-justificationist theory of rationality. tric celestial mechanics in spite of the abundant
\1oreover, how can we ever recognize an episte- evidence against the rotation of the Earth; or Bohr
'Tlological disaster? We have no means to judge and his disciples accepted a theory of light emis-
.\hether the verisimilitude of our successive theo- sion in spite of the fact that it ran counter to Max-
·1es increases or decreases. 52 At this stage we have well's well-corroborated theory.
riot yet developed a general theory of criticism lndeed, it is not difficult to see at least two
even for scientific theories, let al one for theories of crucial characteristics common to bath dogmatic
·ationality 53 : therefore, if we want to falsify our and our methodological falsificationism which are
'Tlethodological falsificationism, we have to do it clearly dissonant with the actual history of sci-
before having a theory of how to do it. ence: that (1) a test is-or must be made-a two-
If we look at history of science, if we try to see cornered fight between theory and experiment sa
1ow some of the most celebrated falsifications that in the final confrontation on/y these two face
happened, we have to corne to the conclusion that each other; and (2) the on/y interesting outcome of
either some of them are plainly irrational or that such confrontation is (conclusive) falsification:
they rest on rationality principles radically differ- "[the on/y genuine] discoveries are refutations of
ent from the ones we just discussed. First of al/, scientific hypotheses."55 However, history of sci-
our falsificationist must deplore the fact that stub- ence suggests that (1 ') tests are-at least-three-
born theoreticians frequently challenge experi- cornered fights between rival theories and experi-
mental verdicts and have them reversed. ln the ment and (2') some of the most interesting
ialsificationist conception of scientific "law and experiments result, prima facie, in confirmation
order" we have described there is no place for rather than falsification.
such successful appeals. Further difficulties arise But if-as seems ta be the case-the history of
from the falsification of theories to which a ceteris science does not bear out our theory of scientific
oaribus clause is appended. 54 Their falsification as rationality, we have two alternatives. One alterna-
tive is to abandon efforts to give a rational expla-
nation of the success of science. Scientific method
(or "logic of discovery"), conceived as the disci-
52 1am using here "verisimilitude·· in Popper's sense: the differ·
pline of rational appraisal of scientific theories-
ence between the truth content and falsity content of a theory.
For the risks involved in estimatrng it, cf my Il %8a], especially and of criteria of progress-vanishes. We, may, of
pp. 395 fi. course, still try to explain changes in "paradigms"
531 tried to develop such a general theory of criticism in my in terms of social psychology. . . . This is
11970]
54 The falsification of theories depends on the high degree of

corroboration of the ceteris paribu1 clause. This, however, is

not always the case. This is whv the methodological falsrfica·
tionist may advise us to rely on our "scientific rnstinct" (Popper
[1934], Section 18, footnote 2) or "hunch" (Brarthwaite [ 1953], 55
Agassi [1959]; he calls Popper's idea of science "scientra
p. 20). negativa" (Agassi [1968]).
 182 Part 3 The Validation of Scientific Knowledge

Polanyi's and Kuhn's way. 56 The other alternative able in the light of, or even forbidden, by T; 58 (2)
1
is to try at least to reduce the conventional ele- T' explains the previous success of T, that is, ail
ment in falsificationism (we cannot possibly elimi- the unrefuted content of T is contained (within the
1 nate it) and replace the naïve versions of method- limits of observational error) in the content of T';
ological falsificationism-characterized by the and (3) some of the excess content of T' is corrob-
theses (1) and (2) above-by a sophisticated ver- orated.59
sion which would give a new rationale of falsifica- ln order to be able to appraise these definitions
tion and thereby rescue methodology and the idea we need to understand their problem background
of scientific progress. This is Popper's way, and and their consequences. First, we have to remem-
the one 1 intend follow. ber the conventionalists' methodological discov-
ery that no experimental result can ever kill a
Sophisticated versus Naïve Methodological theory: any theory can be saved from coun-
Falsificationism. Progressive and terinstances either by some auxiliary hypothesis or
Degenerating Problemshifts. by a suitable reinterpretation of its terms. Naive
Sophisticated falsificationism differs from naive falsificationists solved this problem by relegat-
falsificationism both in its rules of acceptance (or ing-in crucial contexts-the auxiliary hvpothe-
"demarcation criterion") and its rules of falsifica- ses to the realm of unproblematic background
tion or elimination. For the naive falsificationist knowledge, eliminating them from the deductive
any theory which can be interpreted as experi- model of the test-situation and thereby forcing the
mentally falsifiable is "acceptable" or "scien- chosen theory into logical isolation, in which it
tific."57 For the sophisticated falsificationist a the- becomes a sitting target for the attack of test-exper-
ory is "acceptable" or "scientific" only if it has iments. But since this procedure did not offer a
corroborated excess empirical content over its suitable guide for a rational reconstruction of the
predecessor (or rival), that is, only if it leads to the history of science, we may just as well completely
discovery of novel facts. This condition can be rethink our approach. Why aim at falsification at
analysed into two clauses: that the new theory has any price? Why not rather impose certain stan-
excess empirical content (" acceptability" 1) and dards on the theoretical adjustments by which one
that some of this excess content is verified ("ac- is allowed to save a theory? Jndeed, some such
ceptability" 2). The first clause can be checked in- standards have been well known for centuries,
stantly . . . by a priori logical analysis; the sec- and we find them expressed in age-old wisecracks
ond can be checked only empirically and this may against ad hoc explanations, empty prevarica-
take an indefinite time. tions, face-saving, linguistic tricks. 60 We have al-
Again, for the naive falsificationist a theory is ready seen that Duhem adumbrated such stan-
falsified by a ("fortified" . . . ) "observational"
statement which conflicts with it (or rather, which
he decides to interpret as conflicting with it). The 58
1 use "prediction" in a wide sense that includes "postdt( ·
sophisticated falsificationist regards a scientific tion."
59
theory Tas falsified if and only if another theory T' For a detailed discussion oi these acceptance and re1ect1on
ru/ps and tor references to Popper\ work, cf. my 11 Y68al.
has been proposed with the following characteris- pp. l75-YO. ~or some qual1f1cat1ons (concerning continuity
tics: (1) T' has excess empirical content over T; and consistency as regulat1ve principle-.), cf. be/ow, pp. 190-
191.
that is, it predicts nove/ facts, that is, facts improb-
60
Molière, for instance, ridiculed the doctors of h1s Malade
Imaginaire, who offered the vtrtus dormitiva of opium as thP
answer to the question as to why opium produced sleep. One
56 might even argue that Newton's iamous dictum hypothp;es non
Feyerabend, who contributed probably more than anybody fingo was really d1rected against ad hoc explanations-like h1'
else to the spread of Popper's ideas, seems now to have joined own explanation of gravitational forces by an aether-model in
the enemy camp. Cf. his intriguing [1970). order to meet Cartesian ob1ections.
7
5 Ct. above, p. 179. 61
Cf. above, p. 177.
:

1
 Falsification and the Methodology of Scientific Research Programmes 183

dards in terms of "simplicity" and "good sense." 61 Now we can easily understand why we formu-
But when does Jack of "simplicity" in the protec- lated the criteria of acceptance and rejection of
tive belt of theoretical adjustments reach the point sophisticated methodological falsificationism as
at which the theory must be abandoned? 62 ln what we did. 66 But it may be worth while to reformulate
sense was Copernican theory, for instance, "sim- them slightly, couching them explicitly in terms of
pler" than Ptolemaic? 63 The vague notion of Du- series of theories.
hemian "simplicity" leaves, as the naive falsifica- Let us take a series of theories, T1, T2 , T3 , • . •
tionist correctly argued, the decision very much to where each subsequent theory results from adding
taste and fashion. 64 auxiliary clauses to (or from semantical reinterpre-
Can one improve on Duhem's approach? Pop- tations of) the previous theory in order to accom-
per did. His solution-a sophisticated version of modate some anomaly, each theory having at
methodological falsificationism-is more objec- least as much content as the unrefuted content of
tive and more rigorous. Popper agrees with the its predecessor. Let us say that such a series oi
~onventionalists that theories and factual proposi- theories is theoretical/y progressive (or "consti-
'.1ons can always be harmonized with the help of tutes a theoretical/y progressive prob/emshift") if
auxiliary hypotheses: He agrees that the problem each new theory has some excess empirical con-
1s how to demarcate between scientific and pseu- tent over its predecessor; that is, if it predicts some
doscientific adjustments, between rational and ir- novel, hitherto unexpected fact. Let us say that a
rational changes of theory. According to Popper, theoretically progressive series of theories is also
saving a theory with the help of auxiliary hypothe- empirically progressive (or "constitutes an empiri-
ses which satisfy certain well-defined conditions cally progressive problemshift") if some of this ex-
represents scientific progress; but saving a theory cess empirical content is also corroborated, that
with the help of auxiliary hypotheses which do is, if each new theory leads us to the actual discov-
not, represents degeneration. Popper calls such ery of some new fact. 67 Finally, let us call a pro-
:nadmissible auxiliary hypotheses ad hoc hypoth- blemshift progressive if it is bath theoretically and
eses, mere linguistic devices, "conventionalist empirically progressive, and degenerating if it is
stratagems." 65 But then any scientific theory has to not.6B We "accept" problemshifts as "scientific"
be appraised together with its auxiliary hypothe- only if they are at least theoretically progressive; if
ses, initial conditions, etc., and, especially, to- they are not, we "reject" them as "pseudoscien-
gether with its predecessors so that we may see by tific." Progress is measured by the degree to which
what sort of change it was brought about. Then, of a problemshift is progressive, by the degree to
course, what we appraise is a series of theories which the series of theories leads us to the discov-
rather than isolated theories. ery of navel facts. We regard a theory in the series

62 lncidentally, Duhem agreed with Bernard that experiments 66 Cf. above, p. 182.
alone-without simplicity cons1derations-can decide the fate
67 If 1 already know P1 : "Swan A is white," Pw: "Ali swans are
oftheories in physiology. But in physics, he argued, they cannot
([1905) Chapter VI, Section 1) white" represents no progress, because it may only lead to the
63
discovery of such further similar facts as P2 : "Swan Bis white."
Koestler correctly points out that only Galileo created the
So-called "empirical generalizations" constitute no progress. A
myth that the Copernican theory was simple (Koestler [1959),
new fact must be improbable or even impossible in the light of
p. 4761; in fact, "the motion oi the earth [had notl done much
previous knowledge. Cf. above, p. 182.
to simplify the old theories, for though the objectionable
6 8 The appropriateness of the term "problemshift" for a series of
equants had disappeared, the svstem was still bristling with
auxiliary circles" (Dreyer [1906), Chapter XlllJ. theories rather than of problems may be questioned. 1 chose it
partly because 1 have not found a more appropriate alterna-
64 Cf. above, p. 177. tive-"theoryshift" sounds dreadful-partly because theories
65 Popper [1934), sections 19 and 20. 1 have discussed in some are always problematical, they never solve ail the problems
detail-under the heads "monster-barring," "exception-bar- they have set out to salve. Anyway, in the second half of the
ring," "monster-adjustment" -such stratagems as they appear paper, the more natural term "research programme" will re-
in informai, quasi-empirical mathematics; cf. my [1963-4]. place "problemshifts" in the most relevant contexts.
184 Part 3 The Validation of Scientific Knowledge

"falsified" when it is superseded by a theory with the fate of a theory is the result of a test, i.e., an
higher corroborated content. . . . agreement about basic statements." 71 Contrary to
This demarcation between progressive and de- naïve falsificationism, no experiment, experimen-
generating problemshifts sheds new 1ight on the ta/ report, observation statement or we/1-corrobo-
appraisal of scientific-or rather, progressive-ex- rated /ow-level falsifying hypothesis a/one can
p/anations. If we put forward a theory to resolve a lead to falsification. . . There is no falsification
contradiction between a previous theory and a before the emergence of a better theory. 72 But then
counterexample in such a way that the new the- the distinctively negative character of naive falsifi-
ory, instead of offering a content-increasing (sci- cationism vanishes; criticism becomes more diffi-
entific) explanation, only offers a content-decreas- cult, and also positive, constructive. But, of
ing (1 inguistic) reinterpretation, the contradiction course, if falsification depends on the emergence
is resolved in a merely semantical, unscientific of better theories, on the invention of theories
way. A given fact is explained scientifically on/y if which anticipate new facts, then falsification is not
a new fact is a/sa explained with it. . . . simply a relation between a theory and the empiri-
Sophisticated falsificationism thus shifts the cal basis, but a multiple relation between compet-
problem of how to appraise theories to the prob- ing theories, the original "empirical basis," and
lem of how to appraise series of theories. Not an the empirical growth resulting from the competi-
isolated theory, but only a series of theories can be tion. Falsification can thus be said to have a "his-
said to be scientific or unscientific: to apply the torical character." 73 Moreover, some of the theo-
term "scientific" to one single theory is a category ries which bring about falsification are frequently
mistake. 69 proposed after the "counterevidence." This may
The time-honoured empirical criterion for a sound paradoxical for people indoctrinated with
satisfactory theory was agreement with the ob- naïve falsificationism. lndeed, this epistemologi-
served facts. Our empirical criterion for a series of cal theory of the relation between theory and ex-
theories is that it should produce new facts. The periment differs sharply from the epistemological
idea of growth and the concept of empirica/ char- theory of naïve falsificationism. The very term
acter are soldered into one. "counterevidence" has to be abandoned in the
This revised form of methodological falsifica- sense that no experimental result must be inter-
tionism has many new features. First, it denies that preted directly as "counterevidence." If we still
"in the case of a scientific theory, our decision want to retain this time-honoured term, we have
depends upon the results of experiments. If these to redefine it like this: "Counterevidence to T1 " is
confirm the theory, we may accept it until we find a corroborating instance to T2 , which is either in-
a better one. If they contradict the theory, we re- consistent with or independent of T1 (with the pro-
ject it." 70 lt denies that "what ultimately decides viso that T2 is a theory which satisfactorily explains
the empirical success of T1 ). This shows that "cru-

9
b Popper's conflation of "theories" and "series of theories"
prevented him from getting the basic ideas of sophisticated falsi- sequences can be more directly tested by experiment" (ibid; my
ficationism across more successfully. His ambiguous usage led italics).
to such confusing formulations as "Marxism [as the core of a
71
series of theories or of a research programme] is irrefutable" Popper [1934], Section 30.
and, at the same time, "Marxism [as a particular conjunction of 72
"ln most cases we have, before falsifying a hypothesis, an-
this core and some specified auxiliary hypotheses, initial condi- other one up our sleeves" !Popper [1959], p. 87, footnote *Il.
tions and a ceteris paribus clause] has been refuted." (Cf. Pop- But, as our argument shows, we must have one. Or, as Feyera-
per [1963].) bend put it: "The best criticism rs provided by those theories
Of course, there is nothing wrong in saying that an iso- which can replace the rivais they have removed" ([1965], p.
lated, single theory is "scientific" if it represents an advance on p. 2271. He notes that in some cases "alternatives will be quite
its predecessor, as long as one clearly realizes that in this formu- indispensable for the purpose of refutation" (ibid, p. 254). But
lation we appraise the theory as the outcome of-and in the according to our argument refutation without an alternative
context of-a certain historical development. shows nothing but the poverty of our imagination 1n providing a
70
Popper [1945], Vol. Il, p. 233. Popper's more sophisticated rescue hypothesis.
attitude surfaces in the remark that "concrete and practical con- 73
Cf. my [1968a], pp. 387 ff.
 Falsification and the Methodology of Scientific Research Programmes 185
cial counterevidence" -or "crucial experi- tians." Science can grow without any "refuta-
ments"-can be recognized as such among the tions" leading the way. Naïve falsificationists sug-
scores of anomalies only with hindsight, in the gest a linear growth of science, in the sense that
light of some superseding theory. 74 theories are fol lowed by powerfu 1 refutations
Thus the crucial element in falsification is which eliminate them; these refutations in turn are
whether the new theory offers any navel, excess followed by new theories. 77 lt is perfectly possible
1 nformation compared with its predecessor and that theories be put forward "progressively" in
whether some of this excess information is corrob- such a rapid succession that the "refutation" of the
orated. Justificationists valued "confirming" in- nth appears only as the corroboration of the n + 1-
stances of a theory; naive falsificationists stressed th. The problem fever of science is raised by prolif-
refuting" instances; for the methodological falsifi- eration of rival theories rather than counterexam-
cationists it is the-rather rare-corroborating in- ples or anomalies.
stances of the excess information which are the This shows that the slogan of proliferation of
crucial ones; these receive ail the attention. We theories is much more important for sophisticated
are no longer interested in the thousands of trivial than for naive falsificationism. For the naive falsifi-
verifying instances nor in the hundreds of readily cationist science grows through repeated experi-
available anomalies: The few crucial excess-veri- mental overthrow of theories; new rival theories
fying instances are decisive. 75 This consideration proposed before such "overthrows" may speed up
rehabilitates-and reinterprets-the old proverb: growth but are not absolutely necessary . . . ;
Exemplum docet, exempla obscurant. constant prol iferation of theories is optional but
"Falsification" in the sense of naive falsifica- not mandatory. For the sophisticated falsifica-
tionism (corroborated counterevidence) is not a tionist proliferation of theories cannot wait until
sufficient condition for eliminating a specific the- the accepted theories are "refuted" (or until their
ory: ln spite of hundreds of known anomalies we protagonists get into a Kuhnian crisis of confi-
do not regard it as falsified (that is, eliminated) dence).78 While naive falsificationism stresses
until we have a better one. 76 Nor is "falsification" "the urgency of replacing a falsified hypothesis by
in the naive sense necessary for falsification in the a better one," 79 sophisticated falsificationism
sophisticated sense: A progressive problemshift stresses the urgency of replacing any hypothesis
does not have to be interspersed with "refuta- by a better one. Falsification cannot "compel the
theorist to search for a better theory," 80 simply
because falsification cannot precede the better
theory .
. , ln the distorting mirror of naive falsificationism, new theories
which replace old refuted ones are themselves born unrefuted.
The problem-shift from naive to sophisticated
Therefore they do not believe that there is a relevant difference falsificationism involves a semantic difficulty. For
between anomalies and crucial counterevidence. For them,
anomaly is a dishonest euphemism for counterevidence. But in
the naive falsificationist a "refutation" is an exper-
actual history new theories are barn refuted: They 1nherit many imental result which, by force of his decisions, is
anomalies of the old theory. Moreover, frequently it is on/y the
new theory which dramatically predicts that fact which will
made to conflict with the theory under test. But
function as crucial counterevidence against its predecessor, according to sophisticated falsificationism one
white the "old" anomalies may well stay on as "new" anoma-
lies.
must not take such decisions before the al leged
·s Sophisticated falsificationism adumbrates a new theory of
"refuting instance" has become the confirming in-
learning; cf. be/ow, p. 186. stance of a new, better theory. Therefore when-
' 6 lt is clear that the theory T' may have excess corroborated
empirical content over another theory T even if bath T and T'
are refuted. Empirical content has nothing to do with truth or
falsity. Corroborated contents can also be compared irrespec- 77
tive of the refuted content. Thus we may see the rational ity of Cf. Popper [1934], Section 85, p. 279 of the 1959 English
the elimination of Newton's theory in favour of Einstein's, even translation.
78
though Einstein's theory may be said to have been born-like Also cf. Feyerabend [1965], pp. 254-5.
Newton's-"refuted." We have only to remember that "quali- 79
tative confirmation" is a euphemism for "quantitative disconfir- Popper [1959], p. 87, footnote *I.
mation." (Cf. my [1968a]. pp. 384-6.1 80 Popper [1934], Section 30.
186 Part 3 The Validation of Scientific Knowledge

ever we see terms like "refutation," "falsifica- anything) about refuted theories one learns that
tion," "counterexample," we have to check in they are disproved. 82 For the sophisticated falsifi-
each case whether these terms are being applied cationist, learning about a theory is primarily
in virtue of decisions by the na ive or by the sophis- learning which new facts it anticipated; indeed,
ticated falsificationist. 81 for the sort of Popperian empiricism 1 advocate,
the only relevant evidence is the evidence antici-
Sophisticated methodological falsificationism pated by a theory, and empirica/ness (or scientific
offers new standards for intellectual honesty. Justi- characterJ and theoretica/ progress are inseparably
fication ist honesty demanded the acceptance of connected . . . .
only what was proven and the rejection of every- This idea is not entirely new. Leibnitz, for in-
thing unproven. Neojustificationist honesty de- stance, in his famous letter to Conring in 1678,
manded the specification of the probability of any wrote: "lt is the greatest commendation of an hy-
hypothesis in the light of the available empirical pothesis (next to [proven] truth) if by its help pre-
evidence. The honesty of naive falsificationism dictions can be made even about phenomena or
demanded the testing of the falsifiable and the re- experiments not tried." 83 Leibnitz's view was
jection of the unfalsifiable and the falsified. Fi- widely accepted by scientists. But since all ap-
nal ly, the honesty of sophisticated falsificationism praisal of a scientific theory was before Popper
demanded that one should try to look at things appraisal of its degree of justification, this position
from different points of view, to put forward new was regarded by some logicians as untenable.
theories which anticipate novel facts, and to reject Mill, for instance, complains in 1843 in horror
theories which have been superseded by more that "it seems to be thought that an hypothesis .
powerful ones. . . is entitled to a more favourable reception, if
Sophisticated methodological falsificationism besides accounting for ail the facts previously
blends several different traditions. From the empir- known, it has led to the anticipation and predic-
icists it has inherited the determination to learn tion of others which experience afterwards veri-
primarily from experience. From the Kantians it fied."84 Mill had a point; this appraisal was in
has taken the activist approach to the theory of conflict both with justificationism and with proba-
knowledge. From the conventionalists it has bilism: Why should an event prove more, if it was
learned the importance of decisions in method- anticipated by the theory than if it was known
ology. already before? As long as proof was the only cri-
1 should like to emphasize here a further dis- terion of the scientific character of a theory, Leib-
tinctive feature of sophisticated methodological nitz's criterion could only be regarded as irrele-
empiricism: the crucial role of excess corrobora- vant.85 Also, the probability of a theory given
tion. For the inductivist, learning about a new the-
ory is learning how much confirming evidence
supports it; about refuted theories one learns noth-
ing (learning, after all, is to build up proven or
82
probable knowledge). For the dogmatic falsifica- For a defense of this theory of "learning from experience," cf.
Agassi [ 1969[.
tionist, learning about a theory is learning whether 83
Cf. Leibnitz [ 1678[. The expression in brackets shows that
it is refuted or not; about confirmed theories one Leibnitz regarded this criterion as second best and thought that
learns nothing (one cannot prove or probabilify the best theories are those which are proved. Thus Leibnitz's
position-like Whewell's-is a far cry from fui/y fledged so-
phisticated falsification:sm.
84 Mill [1843]. vol. Il, p. 23.
Possibly it would be better in future to abandon these 81
8l . This was J. S. Mill's argument (ibid). He directed it against
terms a/together, just as we have abandoned terms like "induc- Whewel/, who thought that "consilience of inductions" or suc-
tive (or experimental) proof." Then we may call (naive) "refuta- cessful prediction of improbable events verifies (that is, proves)
tions" anomalies, and (sophisticatedly) "falsified" theories "su- a theory. (Whewell [1858]. pp. 95-6.) No doubt, the basic
perseded" ones. Our "ordinary" language is impregnated not contradiction both in Whewell's and in Duhem's philosophy of
only by "inductivist" but also by falsificationist dogmatism. A science is their conflation of heuristic power and proven truth.
reform is overdue. Popper separated the two.
 Falsification and the Methodology of Scientific Research Programmes 187

evidence cannot possibly be influenced, as Key- Let us finally consider how much convention-
nes pointed out, by when the evidence was pro- alism remains in sophisticated falsificationism.
duced; the probability of a theory given evidence Certainly Jess than in naive falsificationism. We
can depend only on the theory and the evi- need fewer methodological decisions. The
dence, 86 and not upon whether the evidence was "fourth-type decision"which was essential for the
produced before or after the theory. naive version 88 has become completely redun-
ln spite of this convincing justificationist criti- dant. To show this we only have to realize that if a
cism, the criterion survived among some of the scientific theory, consisting of some "laws of na-
best scientists, since it formulated their strong dis- ture," initial conditions, auxiliary theories (but
1ike of merely ad hoc explanations, which without a ceteris paribus clause) conflicts with
"though [they] truly express the facts [they set out some factual propositions we do not have to de-
to explain, are] not born out by any other phe- cide which-explicit or "hidden" -part to re-
nomena. " 87 place. We may try to replace any part and only
But it was only Popper who recognized that when we have hit on an explanation of the anom-
the prima facie inconsistency between the few aly with the help of some content-increasing
odd, casual remarks against ad hoc hypotheses on change (or auxiliary hypothesis), and nature cor-
the one hand and the huge edifice of justifica- roborates it, do we move on to eliminate the "re-
tionist philosophy of knowledge must be solved by futed" complex. Thus sophisticated falsification is
demolishing justificationism and by introducing a slower but possibly safer process than naive fal-
new, non-justificationist criteria for appraising sci- sification.
entific theories based on anti-adhocness. Let us take an example. Let us assume that the
Let us look at a few examples. Einstein's theory course of a planet differs from the one predicted.
is not better than Newton's because Newton's the- Sorne conclude that this refutes the dynamics and
ory was "refuted" but Einstein's was not: There gravitational theory applied: The initial conditions
are many known "anomalies" to Einsteinian the- and the ceteris paribus clause have been inge-
ory. Einstein's theory is better than-that is, repre- niously corroborated. Other conclude that this re-
sents progress compared with-Newton's theory futes the initial conditions used in the calcula-
anno 1916 (that is, Newton's laws of dynamics, tions: Dynamics and gravitational theory have
law of gravitation, the known set of initial condi- been superbly corroborated in the last two hun-
tions, "minus" the list of known anomalies such dred years and ail suggestions concerning further
as Mercury' s peri hel ion) because it expiai ned factors in play failed. Yet others conclude that this
everything that Newton's theory had successfully refutes the underlying assumption that there were
explained, and it explained also to some extent no other factors in play except for those which
some known anomalies and, in addition, forbade were taken into account: These people may possi-
events like transmission of light along straight lines bly be motivated by the metaphysical principle
near large masses about which Newton's theory that any explanation is only approximative be-
had said nothing but which had been permit- cause of the infinite complexity of the factors in-
ted by other well-corroborated scientific theories volved in determining any single event. Should we
of the day; moreover, at least some of the unex- praise the first type as "critical," scold the second
pected excess Einsteinian content was in fact cor- type as "hack," and condemn the third as "apo/o-
roborated (for instance, by the eclipse experi- getic"? No. We do not need to draw any conclu-
ments) . . . . sions about such "refutation." We never reject a
specific theory simply by fiat. If we have an incon-
sistency like the one mentioned, we do not have
86
Keynes [19211, p. 305. But cf. my [1968aJ, p. 394.
87 This is Whewell's critical comment on an ad hoc auxiliary
hypothesis in Newton's theory of light (Whewell [1858], Vol. Il,
p. 317.) 88 Cf. above, p. 180.
188 Part 3 The Validation of Scientific Knowledge

to decide which ingredients of the theory we re- deduce-in the same deductive model-further
gard as problematic and which ones as unprob- consequences from the basic statement either with
lematic: We regard ail ingredients as problematic the help of the theory under test or some other
in the light of the conflicting accepted basic state- theory which we regard as unproblematic. Al-
ment and try to replace ail of them. If we succeed though this procedure "has no natural end," we
in replacing some ingredient in a "progressive" always corne to a point when there is no further
way (that is, the replacement has more corrobo- disagreement. 93
rated empirical content than the original), we cail But when the theoretician appeals against the
it "falsified." . . . verdict of the experimentalist, the appeal court
The first, second, and third type decisions of does not normaily cross-question the basic state-
naïve falsificationism 89 however, cannot be ment directly but rather questions the interpreta-
avoided, but as we shail show, the conventional tive theory in the light of which its truth-value had
element in the second decision-and also in the been established.
third-can be slightly reduced. We cannot avoid One typical example of a series of successful
the decision which sort of propositions should be appeals is the Proutians' fight against unfavourable
the "observational" ones and which the "theoreti- experimental evidence from 1815 to 1911. For
cal" ones. We cannot avoid either the decision decades Prout's theory T ("that ail atoms are com-
about the truth-value of some "observational pounds of hydrogen atoms and thus 'atomic
propositions." These decisions are vital for the de- weights' of ail chemical elements must be express-
cision whether a problemshift is empiricaily pro- ible as whole numbers") and falsifying "observa-
gressive or degenerating. 90 But the sophisticated tional" hypotheses, like Stas's "refutation" R ("the
falsificationist may at least mitigate the arbitrari- atomic weight of chlorine is 35-5'') confronted
ness of this second decision by ailowing for an each other. As we know, in the end T prevailed
appeal procedure. over R. 94
Naïve falsificationists do not lay down any The first stage of any serious criticism of a sci-
such appeal procedure. They accepta basic state- entific theory is to reconstruct, improve, its logical
ment if it is backed up by a weil-corroborated fal- deductive articulation. Let us do this in the case of
sifying hypothesis, 91 and let it overrule the theory Prout's theory vis à vis Stas's refutation. First of ail,
under test-even though they are weil aware of we have to realize that in the formulation we just
the risk. 92 But there is no reason why we should quoted, T and R were not inconsistent (Physicists
not regard a falsifying hypothesis-and the basic rarely articulate their theories sufficiently to be
statement it supports-as being just as problem- pinned down and caught by the critic.) ln order to
atic as a falsified hypothesis. Now how exactly show them upas inconsistent we have to put them
can we expose the problematicality of a basic in the foilowing form. T: "the atomic weight of ail
statement? On what grounds can the protagonists pure (homogeneous) chemicai elements are multi-
of the "falsified" theory appeal and win? ples of the atomic weight of hydrogen," and R:
Sorne people may say that we might go on "chlorine is a pure (homogeneous) chemical ele-
testing the basic statement (or the falsifying hy- ment and its atomic weight is 35·5." The last state-
pothesis) "by their deductive consequences" until ment is in the form of a falsifying hypothesis
agreement is finaily reached. ln this testing we

93
This is argued in Popper [1934], Section 29.
s9 Cf. above, pp. 178 and 179. 94
Agassi claims that this example shows that we may "stick ta
90 Cf. above, p. 183. the hypothesis in the face of known facts in the hope that the
facts will adjust themselves ta theory rather than the other way
91 Popper [1934), Section 22.
round" [1966], p. 18). But how can facts "adjust themselves"?
92Cf. e.g., Popper [1959], p. 107, footnote *2. Also cf. above, Under which particular conditions should the theory win1
pp. 180- 181. Agassi gives no answer.
 Falsification and the Methodology of Scientific Research Programmes 189

which, if well corroborated, would allow us to use the pluralistic model the clash is not "between
basic statements of the form 8: "Chlorine X is a theories and facts" but between two high-level
pure (homogeneous) chemical element and its theories: between an interpretative theory to pro-
atomic weight is 35-5'' -where X is the proper vide the facts and an explanatory theory to explain
name of a "piece" of ch lori ne determined, say, by them; and the interpretative theory may be on
its space-time co-ordinates. quite as high a level as the explanatory theory. The
But how well-corroborated is R? The first com- clash is then not any more between a logically
ponent of it says that R1 : "Ch lori ne X is a pure higher-level theory and a lower-level falsifying hy-
chemical element." This was the verdict of the pothesis. The problem should not be put in terms
experimental chemist after a rigorous application of whether a "refutation" is real or not. The prob-
of the "experimental techniques" of the day. lem is how to repair an inconsistency between the
Let us have a closer look at the fine-structure of "explanatory theory" under test and the-explicit
R1 • ln fact R, stands for a conjunction of two or hidden-"interpretative" theories; or, if you
longer statements, T, and T2 • The first statement, wish, the problem is which theory ta consider as
T,, cou Id be this: "If seventeen chemical purifying the interpretative one which provides the "hard"
procedures P,, P2 . . . P, 7 are applied to agas, facts and which the explanatory one which "tenta-
what remains will be pure chlorine." T2 is then: tive/y" explains them. ln a mono-theoretical
"X was subjected to the seventeen procedures P,, model we regard the higher-level theory as an ex-
P2 . . • P17 ." The careful "experimenter" care- planatory theory ta be judged by the "facts" deliv-
fully applied ail seventeen procedures: T2 is to be ered from outside (by the authoritative experimen-
accepted. But the conclusion that therefore what talist): ln the case of a clash we reject the
remained must be pure chlorine is a "hard fact" explanation. 95 ln a pluralistic model we may de-
only in virtue of T,. The experimentalist, while cide, alternatively, to regard the higher-level the-
testing T, applied T,. He interpreted what he saw ory as an interpretative theory ta judge the "facts"
in the light of T,: The result was R1 • Yet in the delivered from outside: ln case of a clash we may
monotheoretical mode/ of the explanatory theory reject the "facts" as "monsters." ln a pluralistic
under test this interpretative theory does not ap- model of testing, several theories-more or less
pear at al/. deductively organized-are soldered together.
But what if T,, the i nterpretative theory, is This argument alone would be enough to
false? Why not "apply" T rather than T, and daim show the correctness of the conclusion, which
that atomic weights must be whole numbers? Then we drew from a different earlier argument, that ex-
this will be a "hard fact" in the light of T, and T, periments do not simply overthrow theories,
will be overthrown. Perhaps additional new puri- that no theory forbids a state of affairs specifia-
fying procedures must be invented and applied. ble in advance. 96 lt is not that we propose a
The problem is then not when we should stick theory and Nature may shout No; rather, we pro-
to a "theory" in the face of "known facts" and pose a maze of theories, and Nature may shout
when the other way round. The problem is not INCONSISTENT. . . .
what to do when "theories" clash with "facts." The problem is then shifted from the old prob-
Such a "clash" is only suggested by the
"monotheoretica/ deductive mode/." Whether a
proposition is a "fact'' or a "theory" in the context 95 The decision to use some monotheoretical mode! is clearly
of a test-situation depends on our methodological vital for the naive falsificationist to enable him to reject a theory
on the sole ground of experimental evidence. /t is in fine with
decision. "Empirical basis of a theory" is a mono- the necessity for him to divide sharply, at /east in a test-situation,
theoretical notion, it is relative to some mono- the body of science into two: the problematic and the unprob-
lematic. (Cf. above p. 178-1791 /tison/y the theory he decides
theoretical deductive structure. We may use it as to regard as problematic which he articu/ates in his deductive
first approximation; but in case of "appeal" by the mode/ of rriticism.
theoretician, we must use a pluralistic mode/. ln 96 Cf. above, p. 174.
190 Part 3 The Validation of Scientific Knowledge

lem of replacing a theory refuted by "facts" to the must take a decision about the acceptance or re-
new problem of how to resolve inconsistencies jection of basic statements. But then we have only
between closely associated theories. Which of the postponed-and possibly improved-the deci-
mutually inconsistent theories should be elimi- sion, not avoided it. 100 The difficulties concerning
nated? The sophisticated falsificationist can an- the empirical basis which confronted "naive"
swer that question easily: One has to try to re- falsificationism cannot be avoided by "sophisti-
place first one, then the other, then possibly both, cated" falsificationism either. Even if we regard a
and opt for that new set-up which provides the theory as "factual," that is, if our slow-moving
biggest increase in corroborated content, which and limited imagination cannot offer an alterna-
provides the most progressive problemsh ift. 97 tive to it (as Feyerabend used to put it), we have to
Thus we have established an appeal procedure make, at least occasionally and temporarily, deci-
in case the theoretician wishes to question the sions about its truth-value. Even then, experience
negative verdict of the experimentalist. The theo- still remains, in an important sense, the "impartial
retician may demand that the experimental ist arbiter" 101 of scientific controversy. We cannot get
specify his "interpretative theory," 98 and he may rid of the problem of the "empirical basis," if we
then replace it-to the experimentalist's annoy- want to learn from experience 102 : but we can
ance-by a better one in the light of which his make our learning less dogmatic-but also less
originally "refuted" theory may receive positive fast and less dramatic. By regarding some observa-
appraisal. 99 tional theories as problematic we may make our
But even this appeal procedure cannot do methodology more flexible: but we cannot articu-
more than postpone the conventional decision. late and include a// "background knowledge" (or
For the verdict of the appeal court is not infallible "background ignorance"?) into our critical deduc-
either. When we decide whether it is the replace- tive model. This process is bound to be piecemeal
ment of the "interpretative" or of the "explana- and some conventional line must be drawn at any
tory" theory that produces novel facts, we again given time.

There is one objection even to the sophisti-

cated version of methodological falsificationism
97 For instance, in our earlier example (cf. above, p. 178 ff.J

some may try ta replace the gravitational theory w1th a new one
which cannot be answered without some conces-
and others may try to replace the radio-optics by a new one: We sion to Duhemian "simplicism." The objection is
choose the way which offers the more spectacular growth, the
more progressive problemshift.
the so-called "tacking paradox." According to our
98 Criticism does not assume a fully articulated deductive struc-
definitions, adding to a theory completely discon-
ture: lt creates it. (lncidentally, this is the main message of my nected low-level hypotheses may constitute a
[1963-4].)
"progressive shift." lt is difficult to eliminate such
99 A classical example of this pattern is Newton's relation to
Flamsteed, the first Astronomer Royal. For instance, Newton
makeshift shifts without demanding that "the addi-
visited Flamsteed on 1 September 1694, when working full time tional assertions must be connected with the con-
on his lunar theory; told him to reinterpret some of his data
since they contradicted his own theory; and he explained to
tradicting assertion more intimately than by mere
him exactly how to do it. Flamsteed obeyed Newton and wrote conjunction." 103 This, of course, is a sort of sim-
to him on 7 October: "Since you went home, 1 examined the
observations 1 employed for determining the greatest equations
of the earth's orbit, and considering the moon's places at the
times. ., 1find that (if, as you intimate, the earth inclines on 100 The same applies to the third type of decision. If we reject a
that side the moon that is) you may abate abt 20" from stochastic hypothesis only for one which, in our sense, super-
it . ." Thus Newton constantly criticized and corrected
sedes it, the exact form of the "rejection rules" becomes /ess
Flamsteed's observational theories. Newton taught Flamsteed,
important.
for instance, a better theory of the refractive power of the atmo-
sphere; Flamsteed accepted this and corrected his original w 1 Popper [1945], Vol. Il, Chapter 23, p. 218.
"data." One can understand the constant humiliation and 102 Agassi is then wrong in his thesis that "observation reports
slowly increasing fury of this great observer, having his data
may be accepted as fa Ise and hence the problem of the empiri-
criticized and improved by a man who, on his own confession,
cal basis is thereby disposed of" (Agassi [19661, p. 20).
made no observations himself: lt was this feeling-1 suspect-
which led finally to a vicious persona! controversv. 103 Feyerabend [ 19651, p. 226.
 Falsification and the Methodology of Scientific Research Programmes 191

plicity requirement which would assure the conti- Popper pointed out, as metaphysical principles. 105
nuity in the series of theories which can be said to For instance, the universa/ anticonventionalist rule
constitute one problemshift. against exception-barring may be stated as the
This leads us to further problems. For one of metaphysical principle: "Nature does not allow
the crucial features of sophisticated falsifica- exceptions." This is why Watkins called such
tionism is that it replaces the concept of theory as rules "influential metaphysics." 106
the basic concept of the logic of discovery by the But what 1 have primarily in mind is not sci-
concept of series of theories. lt is a succession of ence as a whole, but rather particular research
theories and not one given theory which is ap- programmes, such as the one known as "Cartesian
oraised as scientific or pseudo-scientific. But the metaphysics." Cartesian metaphysics, that is, the
members of such series of theories are usual ly mechanistic theory of the universe-according to
connected by a remarkable continuity which which the universe is a huge clockwork (and sys-
welds them into research programmes. This conti- tem of vortices) with push as the only cause of
nuity-reminiscent of Kuhnian "normal sci- motion-functioned as a powerful heuristic prin-
ence" -plays a vital role in the history of science; ciple. lt discouraged work on scientific theories-
the main problems of the logic of discovery cannot like [the "essentialist" version of] Newton's theory
be satisfactorily discussed except in the framework of action at a distance-which were inconsistent
of a methodology of research programmes. with it (negative heuristic). On the other hand, it
encouraged work on auxiliary hypotheses which
might have saved it from apparent counterevi-
A METHODOLOGY OF SCIENTIFIC dence-like Keplerian ellipses (positive heuris-
RESEARCH PROGRAMMES tic).101

1 have discussed the problem of objective ap- Negative Heuristic: The "Hard Core" of
praisal of scientific growth in terms of progressive the Programme.
and degenerating problemshifts in series of scien-
Ali scientific research programmes may be
tific theories. The most important such series in
characterized by their "hard core." The negative
the growth of science are characterized by a cer-
heuristic of the programme forbids us to direct the
tain continuity which connects their members.
modus ta/Jens at this "hard core." lnstead, we
This continuity evolves from a genuine research
must use our ingenuity to articulate or even invent
programme adumbrated at the start. 104 The pro-
"auxiliary hypotheses," which form a protective
gramme consists of methodological rules: Sorne
belt around this core, and we must redirect the
tell us what paths of research to avoid (negative
modus ta/Jens to these. lt is this protective belt of
heuristic), and others what paths to pursue (posi-
auxiliary hypotheses which has to bear the brunt
tive heuristic).
of tests and get adjusted and re-adjusted, or even
Even science as a whole can be regarded as a
completely replaced, to defend the thus-hardened
huge research programme with Popper's supreme
core. A research programme is successful if all this
heuristic rule: "Devise conjectures which have
leads to a progressive problemshift, unsuccessful if
more empirical content than their predecessors."
it leads to a degenerating problemshift.
Such methodological rules may be formulated, as

104 One may point out that the negative and positive heuristic 101 Popper [19341, Sections Il and 70. 1use "metaphysical" as a

gives a rough (implicit) definition of the "conceptual frame- technical term of naive falsificationism: A contingent proposi-
work" (and consequently of the language). The recognition that tion is "metaphysical" if it has no "potential falsifiers."
the history of science is the history of research programmes 106 Watkins [1958].
rather than of theories may therefore be seen as a partial vindi-
107
cation of the view that the history of science is the history of For this Cartesian research programme, cf. Popper [1958]
conceptual frameworks or of scientific languages. and Watkins [ 1958], pp. 350- 1.
192 Part 3 The Validation of Scientific Knowledge

The classical example of a successful research series of "refutations" before ingenious and lucky
programme is Newton's gravitational theory, pos- content-increasing auxiliary hypotheses turn a
sibly the most successful research programme chain of defeats-with hindsight-into a resound-
ever. When it was first produced, it was sub- ing success story, either by revising some false
merged in an ocean of "anomalies" (or, if you "facts" or by adding novel auxiliary hypotheses.
wish, "counterexamples" 108),and opposed by the We may then say that we must require that each
observational theories supporting these anoma- step of a research programme be consistently con-
lies. But Newtonians turned, with brilliant tenacity tent-increasing: that each step constitute a consist-
and ingenuity, one counter-instance after another ent/y progressive theoretica/ problemshift. Ali we
into corroborating instances, primarily by over- need in addition to this is that at least every now
throwing the original observational theories in the and then the increase in content should be seen to
light of which this "contrary evidence" was es- be retrospectively corroborated: The programme
tablished. ln the process they themselves pro- as a whole should also display an intermittent/y
duced new counter-examples which they again progressive empirica/ shift. We do not demand
resolved. They "turned each new difficulty into a that each step produce immediately an observed
new victory of their programme." 109 new fact. Our term intermittent/y gives sufficient
ln Newton's program the negative heuristic rational scope for dogmatic adherence to a pro-
bids us to divert the modus tollens from Newton's gramme in face of prima facie "refutations."
three laws of dynamics and his law of gravitation. The idea of "negative heuristic" of a scientific
This "core" is "irrefutable" by the methodological research programme rationalizes classical con-
decision of its protagonists: Anomalies must lead ventionalism to a considerable extent. We may
to changes only in the "protective" belt of auxil- rationally decide not to allow "refutations" to
iary, "observational" hypotheses and initial con- transmit falsity to the hard core as long as the cor-
ditions.110 roborated empirical content of the protecting belt
1 have given a contrived micro-example of a of auxiliary hypotheses increases. But our ap-
progressive Newtonian problemshift. 111 If we proach differs from Poincaré's justificationist con-
analyse it, it turns out that each successive link in ventionalism in the sense that, unlike Poincaré's,
this exercise predicts some new fact; each step we maintain that if and when the programme
represents an increase in empirical content: The ceases to anticipate novel facts, its hard core
example constitutes a consistent/y progressive the- might have to be abandoned: that is, our hard
oretical shift. Also, each prediction is in the end core, unlike Poincaré's, may crumble under cer-
verified; although on three subsequent occasions tain conditions. ln this sense we side with Duhem
they may have seemed momentarily to be "re- who thought that such a possibility must be al-
futed."112 While "theoretical progress" (in the lowed for; 113 but for Du hem the reason for such
sense here described) may be verified immediately crumbling is purely aesthetic, 114 while for us it is
. . . , "empirical progress" cannot, and in a re- mainly logical and empirical .
search programme we may be frustrated by a long

Positive Heuristic: The Construction of the

108
"Protective Belt" and the Relative Autonomy
For the clarification of the concepts of "counterexample"
and "anomaly" cf. above, p. 180. of Theoretical Science.
109 Laplace [1796], Livre IV, Chapter ii.
Research programmes, besides their negative heu-
110 The actual hard core of a programme does not actually
emerge fully armed like Athene from the head of Zeus. lt de- ristic, are also characterized by their positive heu-
velops slowly, by a long, preliminary process of trial and error. ristic.
ln this paper this process is not discussed.
111
Cf. above, pp. 174-175. For real examples, cf. my [1970].
112The "refutation" was each time successfully diverted to 111
"hidden lemmas"; that is, to lemmas emerging, as it were, from Cf. above, p. 177.
114
the ceteris paribus clause. Ibid.
 Falsification and the Methodology of Scientzfic Research Programmes 193

Even the most rapidly and consistently pro- gravity. This change was not motivated by any
gressive research programmes can digest their observation (the data did not suggest an "anom-
"counter-evidence" only piecemeal: Anomalies aly" here) but by a theoretical difficulty in devel-
are never completely exhausted. But it should not oping the programme. Then he worked out the
be thought that yet unexplained anomalies- programme for more planets as if there were only
"puzzles" as Kuhn might call them-are taken in heliocentric but no interplanetary forces. Then he
random order, and the protective belt built up in worked out the case where the sun and planets
an eclectic fashion, without any preconceived or- were not mass-points but mass-balls. Again, for
der. The order is usually decided in the theoreti- this change he did not need the observation of an
cian's cabinet, independently of the known anom- anomaly; infinite density was forbidden by an (in-
alies. Few theoretical scientists engaged in a articulated) touchstone theory, therefore planets
research programme pay undue attention to "refu- had to be extended. This change involved consid-
tations." They have a long-term research pol icy erable mathematical difficulties, held up New-
which anticipates these refutations. This research ton's work-and delayed the publication of the
policy, or order of research, is set out-in more or Principia by more than a decade. Having solved
less detail-in the positive heuristic of the re- this "puzzle," he started work on spinning balls
search programme. The negative heuristic speci- and their wobbles. Then he admitted interplane-
fies the "hard core" of the programme which is tary forces and started work on perturbations. At
"irrefutable" by the methodological decision of its this point he started to look more anxiously at the
protagonists; the positive heuristic consists of a facts. Many of them were beautifully explained
partially articulated set of suggestions or hints on (qualitatively) by this model, many were not. lt
how to change, develop the "refutable variants" was then that he started to work on bulging plan-
of the research programme, how to modify, so- ets, rather than round planets, etc.
phisticate, the "refutable" protective belt. Newton despised people who, like Hooke,
The positive heuristic of the programme saves stumbled on a first naive model but did not have
the scientist from becoming confused by the the tenacity and ability to develop it into a re-
ocean of anomalies. The positive heuristic sets out search programme, and who thought that a first
a programme which lists a chain of ever more version, a mere aside, constituted a "discovery."
complicated models simulating reality: The scien- He held up publication until his programme had
tist's attention is riveted on building his models achieved a remarkable progressive shift. 116
following instructions which are laid down in the Most, if not all, Newtonian "puzzles," leading
positive part of his programme. He ignores the to a series of new variants superseding each other,
actual counterexamples, the available "data." 115 were foreseeable at the time of Newton's first na-
Newton first worked out his programme for a ive model and no doubt Newton and his col-
planetary system with a fixed point-like sun and leagues did foresee them: Newton must have been
one single point-like planet. lt was in this model
that he derived his inverse square law for Kepler's
ellipse. But this model was forbidden by Newton's
116 Reichenbach, following Cajori, gives a different explanation
own third law of dynamics; therefore, the model of what delayed Newton in the publication of his Principia: "To
had to be replaced by one in which both sun and his disappointment he found that the observational results dis-
agreed with his calculations. Rather than set any theory, how-
planet revolved round their common centre of ever beautiful, before the facts, Newton put the manuscript of
his theory into his drawer. Sorne twenty years later, alter new
measurements of the circumference of the earth had been made
by a French expedition, Newton saw that the figures on which
115 If a scientist (or mathematicianl has a positive heuristic, he
he had based his test were false and that the improved figures
refuses to be drawn into observation. He will "lie down on his agreed with his theoretical calculation. lt was only alter this test
cou ch, shut his eyes and forge! about the data." (Cf. my [ 1963- that he published his law . . . . The story of Newton is one of
4], especially pp. 300 ff., where there is a detailed case study of the most striking illustrations of the method of modern science"
such a programme.) Occasionally, of course, he will ask Nature (Reichenbach [1951], pp. 101-2). Feyerabend criticizes Rei-
a shrewd question: he will then be encouraged by Nature's YES, chenbach's account (Feyerabend [1965], p. 229), but does not
but not discouraged by its NO give an alternative rationa/e.
194 Part 3 The Validation of Scientiftc Knowledge

fully aware of the blatant falsity of his first vari- vide the contact points with reality. Although one
ants.117 Nothing shows the existence of a positive must point out that any "verification" of the n + 1-
heuristic of a research programme clearer than this th version of the programme is a refutation of the
fact: This is why one speaks of "models" in re- nth version, we cannot deny that some defeats of
search programmes. A "mode/" is a set of initial the subsequent versions are always foreseen: lt is
conditions (possibly together with some of the ob- the "verifications" which keep the programme go-
servational theories) which one knows is bound to ing, recalcitrant instances notwithstanding.
be replaced du ring the further development of the We may appraise research programmes, even
programme, and one even knows, more or Jess, after their "elimination," for their heuristic power:
how. This shows once more how irrelevant "refu- How many new facts did they produce, how great
tations" of any specific variant are in a research was "their capacity to explain their refutations in
programme: Their existence is fully expected, the the course of their growth"? 121
positive heuristic is there as the strategy bath for (We may also appraise them for the stimulus
predicting (producing) and digesting them. ln- they gave to mathematics. The real difficulties for
deed, if the positive heuristic is clearly spelt out, the theoretical scientist arise rather from the math-
the difficulties of the programme are mathematical ematica/ difficulties of the programme than from
rather than empirical. 118 anomalies. The greatness of the Newtonian pro-
One may formulate the "positive heuristic" of gramme cornes partly from the development-by
a research programme as a "metaphysical" princi- Newtonians-of classical infinitesimal analysis
ple. For instance, one may formulate Newton's which was a crucial precondition of its success.)
programme like this: "The planets are essentially Thus the methodology of scientific research
gravitating spinning-tops of roughly spherical programmes accounts for the relative autonomy of
shape." This idea was never rigidly maintained: theoretical science: a historical fact whose ration-
The planets are not just gravitational, they have ality cannot be explained by the earlier falsifica-
also, for example, electromagnetic characteristics tionists. Which problems scientists working in
which may influence their motion. Positive heuris- powerful research programmes rationally choose
tic is thus in general more flexible than negative is determined by the positive heuristic of the pro-
heuristic. Moreover, it occasionally happens that gramme rather than by psychologically worrying
when a research programme gets into a degenerat- (or technologically urgent) anomalies. The anom-
ing phase, a little revolution or a creative shift in alies are listed but shoved aside in the hope that
its positive heuristic may push it forward again. 11 Y they will turn, in due course, into corroborations
lt is better therefore to separate the "hard core" of the programme. Only those scientists have to
from the more flexible metaphysical principles ex- rivet their attention on anomalies who are either
pressing the positive heuristic. engaged in trial-and-error exercises . . . or who
Our considerations show that the positive heu- work in a degenerating phase of a research pro-
ristic forges ahead with almost complete disregard gramme when the positive heuristic ran out of
of "refutations": lt may seem that it is the "verifi- steam. (Ali this, of course, must sound repugnant
cations"120 rather than the refutations which pro- to naive falsificationists who hold that once a the-
ory is "refuted" by experiment (by their rule
book), it is irrational (and dishonest) to develop it
117
For a further discussion of Newton's research programme,
further: One has to replace the old "refuted" the-
cf. my [1970]. ory by a new, unrefuted one.) . . .
1 1a For this point cf. Truesdell [1960].
119Soddy's contribution to Prout's programme or Pauli's to
Bohr's (old quantum theory) programme are typical examples of 121
Cf. my [1963-41, pp. 324-30. Unfortunately in 1963-4 1
such creative sh ifts.
had not yet made a clear terminological distinction between
120 theories and research programmes, and this impaired my expo-
A "verification" is a corroboration of excess content in the
expanding programme. But, of course, a "verification" does no! sition of a research programme in informai, quasi-empirical
verify a programme: lt shows only its heuristic power. mathematics. There are fewer such shortcomings in my [19711
 Falsification and the Methodology of Scientific Research Programmes 195

REFERENCES taphysique et de Morale, 7, pp. 375-425, 503-62,

706-31.
Agassi [1959]: "How are Facts Discovered?" Impulse, Le Roy [1901]: "Un Positivisme Nouveau," Revue de
3, No. 10, pp. 2-4. Métaphysique et de Morale, 9, pp. 138-53.
Agassi [1966]: "Sensationalism," Mind, N.S. 75, pp. 1- Medawar [1967]: The Art of the Soluble, 1967.
24. Milhaud [1896]: "La Science Rationnelle," Revue de
Agassi [1968]: "The Novelty of Popper's Philosophy of Métaphysique et de Morale, 4, pp. 280-302.
Science." International Ph1/osophical Quarter/y, 8, Mill [1843]: A System of Logic, Ratiocinative and Induc-
pp. 442-63. tive, Being a Connected View of the Princip/es of Evi-
Agassi [1969]: "Popper on Learning from Experience," dence, and the Methods of Scientific Investigation,
in Rescher (ed.): Studies in the Philosophy of Science, 1843.
1969. Musgrave [1969a]: lmpersona/ Knowledge, Ph.D. The-
Braithwaite [1938]: "The Relevance of Psychology to sis, University of London, 1969.
Logic," Aristotelian Society Supp/ementary Volumes, Musgrave [196%]: Review of Ziman's "Public Knowl-
17, pp. 19-41. edge: An Essay Concerning the Social Dimensions of
Braithwaite [1953]: Scientific Explanat1on, 1953. Science," in The British Journal for the Phi/osophy of
Dryer [1906]: History of the Planetary Systems from Science, 20, pp. 92-4.
Thales to Kepler, 1906. Neurath [1935]: "Pseudorationalismus der Falsifika-
Duhem [1905]: La Théone Physique, Son Obiet et Sa tion," Erkenntnis, 5, pp. 353-65.
Structure, 1905. English translation of the second Poincaré [1891]: "Les géométries non euclidiennes,"
(1914) edition: The Aim and Structure of Physical Revue des Sciences Pures et Appliquées, 2, pp. 769-
Theory, 1954. 74.
Einstein 11949]: "Autobiographical Notes," in Schilpp Poincaré 11902]: La Science et l'Hypothèse, 1902.
(ed.): Albert Einstein, Ph1/osopher-Sc1entist, 1, pp. 2- Popper [1933]: "Ein Kriterium des empirischen Charak-
95. ters theoretischer Systeme," Erkenntnis, 3, pp. 426-
Feyerabend [1965]: "Reply to Criticism," in Cohen and 7.
Wartofsky (eds.): Boston Studies in the Philosophy of Popper [1934]: Logik der Forschung, 1935 (expanded
Science, Il, pp. 223-61. English edition: Popper [1959]).
Feyerabend [1969]: "Problems of Empiricism Il," in Popper [1935]: "lnduktionslogik und Hypothe-
Colodny (ed.): The Nature and Funct1on of Sc1entific senwahrscheinlichkeit," Erkenntnis, 5, pp. 170-2.
Theory, 1969. Popper [1945]: The Open Society and its Enemies, 1-11,
Feyerabend [1970]: "Against Method," Minnesota 1945.
Studies for the Philosophy of Science, 4, 1970. Popper [1957]: "The Aim of Science," Ratio, 1, pp. 24-
Hempel 11952]: "Sorne Theses on Empirical Certainty," 35.
The Review of Metaphysics, 5, pp. 620- 1. Popper [1958]: "Philosophy and Physics"; published in
Keynes [1921]: A Treatise on Probability, 1921. Atti del XII Congresso lnternaziona/e di Fi/osofia, Vol.
Koestler [1959]: The Sleepwalkers, 1959. 2, 1960, pp. 363-74.
Lakatos [1962]: "lnfinite Regress and the Foundatlons of Popper [1959]: The Logic of Scientific Oiscovery, 1959.
Mathematics," Aristote/fan Society Supplementary Popper [1963]: Conjectures and Refutations, 1963.
Volume, 36, pp. 155-84. Popper [1968]: "Remarks on the Problems of Demarca-
Lakatos [1963-4]: "Proofs and Refutations," The British tion and Rationality," in Lakatos and Musgrave (eds.):
journal for the Philosophy of Science, 14, pp. 1-25, Problems in the Philosophy of Science, 1968, pp. 88-
120-39, 221-43, 296-342. 102.
Lakatos [1968a]: "Changes in the Problem of Inductive Reichenbach [1951]: The Rise of Scientific Philosophy,
Logic," in Lakatos (ed.J: The Prob/em of Inductive 1951.
Logic, 1968, pp. 315-417. Russell [1943]: "Reply to Critics," in Schilpp (ed.): The
Lakatos [1968b]: "Criticism and the Methodology of Phi/osophy of Bertrand Russell, 1943, pp. 681-741.
Scientific Research Programmes," in Proceed1ngs of Truesdell [1960]: "The Program toward Rediscovering
the Aristote/ian Society, 69, pp. 149-86. the Rational Mechanics in the Age of Reason," Ar-
Lakatos [ 1970]: The Changing Logic of Scientific Dis- chive of the History of Exact Sciences, 1, pp. 3-36.
covery, 1970. Watkins [1957]: "Between Analytic and Empirical,"
Lakatos [ 1971]: Proofs and Refutations and Other Essays Philosophy, 32, pp. 112-31.
in the Philosophy of Mathematics, 1971. Watkins [ 1958]: "lnfluential and Confirmable Metaphy-
Laplace [1796]: Exposition du Système du Monde, sics," Mind, N.S. 67, pp. 344-65.
1796. Watkins [1960]: "When are Statements Empirical ?"
Leibniz [1678]: LettertoConring, 19.3.1678. British Journal for the Philosophy of Science, 10, pp.
Le Roy [1899]: "Science et Philosophie," Revue de Mé- 287-308.
196 Part 3 The Validation of Scientific Knowledge

Whewell [1837]: History of the Inductive Sciences, the second part of the philosophy of the inductive
from the Earliest to the Present Time. Three volumes, sciences. Third edition, 1858.
1837. Wisdom [1963]: "The Refutability of 'lrrefutable'
Whewell [1840]: Philosophy of the Inductive Sciences, Laws," The British Journal for the Phi/osophy of Sci-
Founded upon their History. Two volumes, 1840. ence, 13, pp. 303-6.
Whewell [1858]: Novum Organon Renovatum. Being