Vous êtes sur la page 1sur 11



Popper's Logik der Forschung ( 1 935)* (see Reichenbach 1935 and Carnap 1935b) contains many remarkable passages whose significance for the logic of science has already been acknowledged by Carnap. But by a certain kind of pseudoratiomlism, Popper blocks his own way to a full appreciation of the practice of research and the history of research to which his book is basically devoted. Namely, he does not use the ambiguity of all factual sciences as the basis of his comments, but, following Laplace's spirit, as it were, aims at one unique distinguished system of statements as the pattern or paradigm of all the factual sciences. One can enter the debate without much preparation because Popper fortunately pursues certain basic ideas that were developed within the Vienna Circle, especially in connection with physicalism, in order to overcome the metaphysics of 'finality'. The basic ideas to which Popper's attitude is, on the whole, close, are approximately these: In logically analysing the factual sciences as masses of statements, our starting point is that we can change all factual statements that are constructed similarly to those of physics, as well as 'protocol statements' under certain circumstances. In the effort to get consistent masses of statements, we discard certain statements, alter others, without, however, being able to start from absolute 'atomic statements' or other conclusive elements.

Although Popper on the whole advocates similar views and thus avoids certain errors, he on the other hand still uses well-defined theories built up of clean statements as models, so to speak, of the factual sciences. Through the form of his 'basic statements' is defined what is to be regarded as an empirical, that is a 'falsifiable', statement (p. 87). Theories, according to him, are tested by basic statements that were acknowledged beforehand for the time being (p. 109). They are rejected if these basic statements "confirm a falsifying hypothesis" (p. 87, 8811). Falsification is the basis of all of Popper's further
Translation of Neurath 1935b [ON2201



comments. His thoughts are constantly circling around a certain ideal; though he does not call it attainable, he uses it somehow as a model when he wants to come to an understanding of what it means that an empirical scientific system runs aground on 'the' experience (p. 41). To this end, according to him,
[an easily falsifiible] theory would describe ' u particular world' as precisely as a or theory can; for it would single out the world of 'our experience' from the class of all logically possible worlds of experience with the greatest precision attainable by theoretical science. All the events or classes of occurrences which we actually encounter and observe, and only these, would be characterized as 'permitted' (p. 113).

Again and again the approximation to this general system plays a part in Popper's comments, as we shall see.

We, on the other hand, try to use models that give no scope at all to thoughts of an ideal of this kind. We start from masses of statements whose connection is only partly systematic, which we also discern only in part. Theories and single communications are placed side by side. While the scholar is working with the help of part of these masses of statements, supplementary additions are made by others, which he is prepared to accept in principle without being quite certain what the logical consequences of this decision might be. The statements from the stock with which one really works use many vague terms, so that 'systems' can always be separated only as abstractions. The statements are linked to each other sometimes more closely, sometimes more loosely. The interlocked whole is not transparent, while systematic deductions are attempted at certain places. This situation is not open to the idea of an 'infinite regress', whereas Popper has to reject it especially in a certain connection (p. 90). If one wants to say that Popper starts from model-systems, one could say that we, on the other hand, start from model-encyclopedias; this would express from the outset that systems of clean statements are not put forward as the basis of our comments.

We believe we are doing the most justice to scientific work if, in our model construction, we set out from the assumption that always the whole mass of statements and all methods can come under discussion.



Certainly we demand of an empiricist that he accept only encyclopedias within which predictions must conform with protocol statements; by our work we can also be led to alter slightly the form of the protocol statements. For while the form of the protocol statements may be more or less established beforehand, the individual protocol statements that are characteristic for a certain encyclopedia and function as test-statements, are not previously distinguished. For the purpose of the model discussion one should realise that for some scientific work, use is made of one of the various encyclopedias that are considered consistent. By thus accepting one definite encyclopedia, definite theories, hypotheses, predictions and their test-statements have been accepted as well. Various factors determine the methodical scientist in his choice of a model. We deny that the encyclopedia preferred by the scientist can be logically selected by using a method that can be only generally outlined. Together with this we not only deny that there could be general methods of 'induction' for the factual sciences, but also that there could be general methods of 'testing' - however, Popper advocates just such general methods of 'testing'. In our way of viewing things 'induction' and 'testing' are linked much more closely than in Popper's. Though we reject a model of science as being a closed system with such general methods, we are still of the opinion that every presentation of scientific research must attempt to present the methods applied in as explicit detail as possible and above all to give adequate credit to each formation of theoretical systems within an encyclopedia. It may happen that certain of Popper's trains of thought that claim the greatest generality, have some special value for special problems of research, within a narrower framework of the kind we have hinted at. In his attack on Reichenbach's works, Popper himself seems to overlook entirely the fact that in spite of their tendency to establish a general theory of induction, they obviously are valuable for scientific research within a more limited space.

Whereas Popper does not want to treat 'induction', this "unfounded anticipation" logico-systematically, not even in its special forms, he tries to characterise falsification logically, as a general method, as strictly as possible - though he must admit that this cannot be done precisely - and to base the whole logic of scientific research uniformly on it. When Popper replaces 'verification' by 'confirmation' of a theory, we replace 'falsification' by 'shaking' of a theory. When a scientist has chosen a



certain encyclopedia (mostly characterised by certain rather general theories that are missing in other encyclopedias available) he will not be induced at once by any negative results to sacrifice a theory, but he will first give careful thought to what the encyclopedia, which he would give up together with the theory, might have been able to achieve for him in the future. Negative results can shake his confidence in an encyclopedia, but not reduce it automatically to zero so to speak through the application of certain rules. We can very well imagine that a falsifying hypothesis that Popper would call 'confirmed' is pushed aside by a successful scientist because, on the basis of very serious general considerations, he deems it an impediment to the development of science that itself would show how this objection is to be refuted. Such a decision may be difficult to make; it is certainly not supported by Popper's tendency always to envisage sections as falsifying entities, and not the total encyclopedia. When a traditional total view is threatened, Popper's stand is, in principle as it were, on the side of the aggressor. It would be very interesting to show what the defensive motions of the practitioners are in such cases. The practitioners of research are at first especially seriously disturbed by such a change. Popper, however, sees the main resistance not in such practitioners and their general attitude, but among the conventionalists (pp. 41, 79-82, etc.). He sketches a type of conventionalism that is perhaps discussed among school philosophers and may occasionally occur among philosophising theorists, but is hardly characteristic of men engaged in the practice of research. This would have to be discussed in conjunction with the history of the sciences. Moreover the reason why a cautious scientist accepts an encyclopedia with certain theories cannot be generally determined by Popper's 'simplicity' (p. 136), whatever value his comments on this subject may otherwise have (p. 136ff.). The unconditional preference for falsification cannot be successfully maintained in the framework of a theory of research. We put 'shaking' side by side with confirmation and try to present each in its way as explicitly as possible, from case to case.


Since Popper starts from the 'modus tollens' of classical logic as his paradigm (p. 41), he calls 'universal singular statements' (that are the 'indefinite existential statements') 'metaphysical', that is non-empirical statements, because according to him they are not falsifiable (p. 69). However, we see what a



blessing they were in the history of the sciences, and we can design a theory of research in which they play a legitimate part. In order to be able to apply his paradigm with the least possible restriction Popper suggests regarding 'natural laws' as statements of not merely 'numerical', but always of 'specific' generality. We should think that a theory of research should phrase its methods so tolerantly that it can satisfy both scientists who from special caution set up all laws only for a limited sphere, viz. treat the world as finite (this is even mentioned by Popper himself), as well as scientists who for some reason just prefer formulations of specific generality of the kind that Popper envisages. In astronomy, geology, sociology and many other disciplines in which experiments - which are overstressed by Popper - play only a small role, such indefinite existential statements, as one-sidedly decidable predictions, are components of normal research, though less, of course, in optics or acoustics. If, for example, we say that on a future day at a certain place a comet can be observed, we have "an only one-sidedly decidable statement before us. If namely the statement is true, soon the day will come when we can decide that it is true; if, however, it is not true, there will never be a day when we can decide that it is untrue" (Reichenbach 1930, p. 168). How significant can it be that a scientist, for example, searches continuously through a certain area of the sky because, by a confirmation of his prediction of a return of a comet at that place, a perhaps very bold theory would be confirmed anew, whereas no falsification of it in Popper's sense seems to be possible in a foreseeable future. Just as Popper counts these 'universal singular statements' among metaphysics, he is inclined to count models that give no access to immediate falsification among 'metaphysical regions' (p. 277). Popper, for example, subsumes the older corpuscular theory of light under 'metaphysical ideas', whereas we would certainly subsume it under the series of scientific models, because in a vague way it shows that certain correlations of optical phenomena, for example, with which we are acquainted from our encyclopedia without special theoretical connections could, according to their type, be deduced from certain more general premises, for example, a corpuscular theory. In our view there are many intermediary stages between these somewhat vague models and the more definite ones of our science. For us there is no dividing line that is supposed to exist between 'falsifiable' and 'nonfalsifiable' theories. We only try to discuss the cases of 'confirmation' and 'shaking' as explicitly as possible.



It is not enough for Popper that the statements of the factual sciences are potentially testable according to their form (whether this form can be described precisely may remain undecided), that is 'unmetaphysical' in our view (see especially Camap), but he stresses excessively that they should be actually testable. This is a proposal for restriction that we cannot find recommendable for the theory of scientific research.
Any empirical scientific statement can be presented (by describing experimental arrangements, etc.) in such a way that anyone who has learned the relevant technique can test it (p. 99).

The over-emphasis on 'falsification' makes Popper see the practice of research all too much from the angle that
the [theorist] puts certain definite questions to the experimenter, and the latter, by his experiments, tries to elicit a decisive answer to these questions, and to no others (p. 107).

Collections of data (sky photographs, etc.), travel journals (for example, Darwin's journal of his voyage around the world is most instructive about these problems) must of course set out from certain theoretical attitudes to make a selection among possible statements feasible, but these theoretical attitudes are not identical with Popper's acute approaches to the theory that are somehow to enforce his 'falsification'. He speaks rather contemptuously of that "myth of a scientific method that starts from observation and experiment and then proceeds to theories. (This legendary method, by the way, still inspires some of the newer sciences which try to practice it because of the prevalent belief that it is the method of experimental physics.)" (p. 279) How much ethnographic material has often to be accumulated before a theory is reached, and how often a group of events is systematically described in physics before it can be arranged. I remember the voluminous literature on 'magnetism of rotation' in the twenties of the nineteenth century. Precise data were available on the basis of which predictions could be made, how, for example, a magnetic needle would move if a copper disc is rotated; but there was no mention of the incorporation of these formulations into a more general theory. How much of the voluminous observation material that was assembled in the fight against the elementary electrical quantum mentioned by Popper - can perhaps fit in later theoretically; for the time being a great many observation statements that seem to contradict the



theory of the elementary quantum are not regarded as essential 'shakings' because one just deems the 'confirmations' of the theory of the elementary quantum to be very significant. Popper, however, wishes to see forceful decisions forcefully founded. That is probably a basic tendency of many pseudorationalist endeavours that should be explained perhaps with the help of the 'psychology of decision'. People who carry out one definite action on the basis of one definite decision are often not content to have carried out such a decision after weighing many individual factors; if they are not able to receive 'transcendental' approval, they at least would like to be able to refer to an unambiguous logical deduction as justification. Whereas people of our attitude sometimes waver between two decisions - should we regard something as a serious shaking or should we simply disregard it for the time being - Popper's formulations obviously point to a more absolutist posture: "But if the decision is negative, or in other words, if the conclusions have been falsified, then their falsification also falsifies the theory from which they were logically deduced" (p. 33). - as if there were a system that could be so cleanly unveiled that such a procedure was possible. Understandably Popper, with such an attitude, must overestimate the usability of the concept 'degree of falsifiability' (pp. 118-119) for the analysis of the work of research. Out of this whole attitude one can probably explain why Popper - in spite of all the warnings by Duhem - likes so much to speak of the 'experimenturn crucis' (pp. 246,277, also p. 237ff.):
In general we regard an inter-subjectively testable falsification as final (provided it is well tested): this is the way in which the asymmetry between verification and falsification of theories makes itself felt. Each of these methodological points contributes in its own peculiar way to the historical development of science as a process of step by step approximations (p. 268).

W have already expressed doubts about these "step-by-step approximations" e and shall have to say more about them. Popper's opinion is, for example, that 'occult effects' have not to be taken seriously because they cannot be reproduced at any time (p. 45). In reply it should be pointed out that there are many non-reproducible but well-attested effects that are theoretically safeguarded and are taken very seriously. However, there is no recognisable advance in 'occult' research (a fact to which Philipp Frank occasionally referred); it has often come about through deception, etc. These are arguinents, however, that are not derived from the overemphasis on experiments for which Popper has a liking. We could sketch the model of a development of science that does not recognise any experiments, for example, along



the lines of Plato's parable of the cave; he tells of prisoners who were chained to the wall and knew perfectly well how to predict shadows and voices although they were deprived of any possibility of making experiments. It is in no way my intention to dismiss the experimental method as of minor significance; but only to reject the idea that the experimental method were as decisive for science as one had to assume from Popper's individual remarks and his total theory of falsification. It is the aim of this essay to fence off certain of Popper's trains of thought that introduce the old philosophical absolutism in a new shape, but not to enter into detailed discussions; otherwise it would be stimulating, in connection with this over-emphasis on reproducible effects, to deal with Popper's remarks on quantum mechanics that distinguish between 'measurement' and 'separation' (p. 238). Nor do we want to deal with Popper's discussion of probability problems (Carnap, Hempel, Reichenbach have already said something about this) though they play a considerable role in his book, for the fundamental conception is not affected by this. It does seem, however, that here too Popper creates difficulties for himself in dealing with certain problems of research, because of his way of putting the questions (p. 195ff.).

In this book, we see Popper's attitude, which is not adapted to empirical research, as a consequence of his decision to choose for his paradigm a system 'that is composed of clean statements and therefore suggests the 'modus tollens'. This sympathy for 'cleanliness' seems to play a part when Popper decidedly rejects our proposal to use 'protocol statements' as teststatements in our encyclopedia model. The protocol statements - in their simiplified form: "Karl's protocol: (in the room is a table perceived by Karl)" - came about as a result of our attempt to avoid a special 'experiential language' ('phenomenal language') and to use nothing but the unified language of physicalism. It is also important to see in this way at once that complex (messy) statements of little cleanliness - 'Ballungen7 -are the basic material of the sciences. Popper is wrong when he thinks that these protocol statements were intended as elementary statements (p. 35). In this form they are even a protest against elementary statements. (Carnap, who on this point, is closer to Popper's proposals, uses the term 'protocol statements' in a sense that differs slightly from the sense in which I use it.) If protocol statements are, in the last resort, the test-statements of the encyclopedia model (that does not mean that one has always to refer to them),



then there is no reason to speak of more or less complex test-statements (p. 126-128). Curiously enough Popper thinks:
Most people would see that any attempt to base logical statements on protocol sentences is a case of psychologism. But curiously enough, when it comes to empirical statements, the same kind of thing goes today by the name of 'physicalism' (pp. 98-99).

Here he overlooks the fact that he himself regards protocol statements as possible, though hardly suitable, basic statements (p. 105). The protocol statements are of a different nature than logical statements; they are indeed statements of the factual sciences; their confrontation with other factual statements at once secures their significance, a confrontation which they do not have with statements of logic. The protocol statements in the form suggested by us have the advantage that they can be maintained if one accepts or rejects the expression within the brackets - taken as an independent statement. If the protocol is accepted - rejection of a protocol is not a frequent occurrence - and in addition the expression within the brackets is taken in isolation, then the protocol can be characterised as a 'reality statement'; if, however, the expression within the brackets taken in isolation is rejected, then the protocol can perhaps be called an 'hallucination statement'. Popper is of the opinion that it is "a widely spread prejudice that the statement, 'I see that the table here is white' has epistemologically greater merits than the statement, 'The table here is white'" (p. 99). For us such protocol statements have the merit of greater stability. The statement: 'In the sixteenth century people saw fiery swords in the sky' can be retained whereas the statement 'There were fiery swords in the sky' would have to be deleted. Just the continuity of forrnulations, however, plays a great part in the selection of model encyclopedias. Such continuity rests in part on constant use of quatemio terminorum; though in contrast with purity this makes possible a connection from people to people, from age to age, from scientist to scientist (problems of this kind are discussed by Ajdukiewicz). When a primitive man says: 'The river runs through the valley,' he certainly defines the terms in a way that is different from that of the European who goes on using the statement. Compared with such impurity, the impurity of protocol statements plays a minor role, though it has to be admitted that the statements of theoretical physics as long as they are not used to formulate predictions tested by protocol statements - give the impression of greater purity. W do not believe that Popper with his attempt to introduce 'observable' e "as an undefined [basic] term which becomes sufficiently precise in use"



(p. 103) and to operate with terms like 'macroscopic', etc., can master the difficulties that result if one wants to turn, for example from the research work of experimental physicists to that of sociologists and psychologists.

In order to be able to design a model for the history of research that reflects its characteristic changes, it is not necessary to take the change of protocol statements into account. It is, however, essential that the stock of successful predictions change. If theory I produces group A of good predictions and theory 11, group A + B of good predictions, we should say that theory I1 is the more successful of the two and say that the stock of predictions A is an approximation to the stock of predictions A +B. However, this in no way implies that the principles of theory I have to be an approximation to the principles of the more successful theory 11. This is logically obvious, but this approximation is not even historically always given. We believe it points to Popper's basic pseudorationalistic attitude when he comments:
For a theory which has been well corroborated can only be superseded by one of a higher level of universality; that is, by a theory which is better testable and which, in addition, contains the old, well-corroborated theory - or at least a good approxirnation to it (p. 276; see also p. 268).

Duhem, whom Popper mentions more than once, shows very beautifully with the various stages of gravitation theory how little they can be seen as 'approximations' to successive stages. Though Popper may declare that science is not "a system which steadily advances towards a state of finality" (p. 278), the above-mentioned passage still indicates that he thinks of this succession of theories when he speaks of "the belief that there are regularities which we can unveil, discover" (see p. 252ff.). These phrases all fit with the basic tendency that we have characterised and that is developed expressly at more than one place. If we want to make a choice among several encyclopedias we permanently use the unified language of physicalism without being forced to use such a terminology, gliding into the metaphysical, which in roundabout ways reintroduces, in the last resort, the term 'real world'.

Popper's pseudorationalistic tendency can be seen historically as a kind of



metaphysical residue from the development of 'philosophy', for this view cannot emerge from the analysis of the factual sciences that are operated without metaphysics. It would be in agreement with this historical supposition that Popper advocates a special 'theory of knowledge' side by side with the logic of science and the factual sciences. Perhaps this closeness to certain metaphysical tendencies explains why Popper behaves to Kant and other metaphysicians much more kindly than to the group of thinkers he calls 'the' positivists - without, however, characterising it sufficiently by an indication of a system of doctrines or an enumeration of names.
The positivist dislikes the idea that there should be meaningful problems outside the field of "positive' empirical science - problems t o be dealt with by a genuine philosophical theory. He dislikes the idea that there should be a genuine theory of knowledge, an epistemology or a methodology. He wishes to see in the alleged philosophical problems mere 'pseudo-problems' or 'puzzles'. . . .Time and again an entirely new philosophical movement arises which finally unmasks the old philosophical problems as pseudoproblems, and which confronts the wicked nonsense of philosophy with the good sense of meaningful, positive, empirical science. And time and again d o the despised defenders of 'traditional philosophy' try to explain to the leaders of the latest positivistic assault that the main problem of philosophy is the critical analysisof the appeal to the authority of 'experience' - precisely that 'experience' which every latest discoverer of positivism is, as ever, artlessly taking for granted (pp. 51-52).

This pleading in favour of traditional [systematic] philosophy makes us foresee that later it will be shown which important role it is called to play as teacher of scientific empiricism, which sees its special fundamental task in the elimination of 'apparent problems'. The pseudorationalism in Popper's view would best make us understand why he could feel attracted by traditional philosophy and its absolutism while his book contains so much of that analytical technique advocated precisely by the Vienna Circle. The aim here was not to give a general presentation of Popper's view, but rather to criticise the absolutism of falsification that is in many ways a counterpart to the absolutism of verification which Popper attacks. It is precisely this book, which is close to the scientific empiricism of the Vienna Circle, that shows once again very clearly that the road to science is far from free of certain residues of compact metaphysics which can only be overcome by common work.

All references in this paper are to the 1968 edition of the English translation.