1 Introduction
The Vienna Circleâs protocol sentence debate is a complex matter encompassing the issues of truth, physicalism, and intersubjective communicability. In this paper I shall confine myself to the core issue of the debate, namely the problem of confirmation. To begin with, it can be stated that protocol sentences are meant to provide direct contact with experience: they are sentences which justify all other synthetic sentences. Therefore, the theory of protocol sentences is intended to clarify the basic contention of the empiricist stance, i.e., the claim that synthetic sentences have to be justified by reference to experience.
Of all the proponents of the debate, surely Carnap is the one whose standpoint is most difficult to grasp. At least on the surface, it is obvious which stance the other players take. According to Neurath, protocol sentences deal â in a somewhat strange interlaced structure â with physical states including states of the observer himself. Popper takes them as singular existential sentences on
By no means these positions are per se free from obscurities and problems. Unfolding these views amounts to no less than far-reaching qualifications of the original idea of the role of protocol sentences, i.e., their justificatory function by reference to experience.1 Nonetheless, Carnap differs in so far that even such an abbreviating and distorting characterization of his position is simply not available. In Carnapâs case, one is inclined to characterize his view as conventionalist. This is not false, conventionalist ideas in one or another form play a major role throughout his whole career. However, this will not suffice to distinguish his view from those of Popper and Neurath, for example.
In the first place, the difficulty of easily grasping Carnapâs position is due to the simple fact that there is no single Carnapian theory of protocol sentences. Contrary to his colleagues or opponents, his views on this issue are âwork in progressâ. Within a relatively short time, from Der logische Aufbau der Welt on,2 his views had developed rather quickly. On the one hand, as I will try to show, this development was driven by the problem of adapting the Aufbau to empiricism. On the other hand, naturally, this development was fueled by the ongoing debate and, most notably, by a kind of pressure from Neurath. Carnapâs development only came to an end in 1936 with the publication of âTestability and Meaningâ. This is quite late, considering that the whole debate is usually presented as taking place in the first half of the '30s. While Carnapâs extensive paper was immediately recognized as being a substantial and path-breaking contribution towards a liberalized empiricism, it received very little attention in the context of the protocol sentence debate.
To secure fruitful discussion, I shall narrow down the subject, i.e., I will start by focusing on the problem of demarcating protocol sentences. Speaking of a justificatory role of protocol sentences presupposes a distinction between protocol sentences and ordinary synthetic sentences.3 But how do we achieve the unambiguous characterization of the former?
- âepistemological: protocol sentences are infallible, incorrigible, etc.
- âsemantical/syntactical: the distinctive feature is to be found either in the content of protocol sentences (e.g., they deal with present experiences) or in their special syntactical form.
- ânaturalistic: protocol sentences are to be distinguished not by considerations a priori, but by the results of science.
- âconventional: it is simply a matter of decision which sentences are used as protocol sentences.
Not being on the same level, there are several possibilities to combine, and in fact, the characterizations of Neurath, Popper and Schlick stated above consist in combinations. It is not only possible to employ some of these features together, but they seem to be related. The epistemological thesis of infallibility, for example, can only be entertained together with the semantical thesis that the content of protocol sentences does not exceed what is momentarily present to the mind.
2 From the Aufbau to âTestability and Meaningâ
2.1 The Aufbau and Its Base
In our system form, the basic elements are to be called experiences of the self after the construction has been carried out; hence, we say: in our constructional system, âmy experiencesâ are the basic elements.6
[â¦] in fact, the constructed objects are objects of conceptual knowledge only qua logical forms which are generated in a certain way. Ultimately, this holds also for the basic elements of the constructional system. [â¦] It is only through this procedure, that is, only as constructed objects, that they become objects of cognition in the proper sense of the word, in particular, objects of psychology.8
How cognition can proceed from one object to another, how, in what sequence, and in which form the levels of a system of cognition can be formulated, â all this is contained in the indicated material. The theory of knowledge cannot ask any further questions.10
It would go beyond the scope of this paper to explore the Aufbau and its essentially neo-Kantian conception of epistemology in more detail,11 nor can we further deal with the problem of the relation of the constructional and epistemological base within the Aufbau.12 It is out of the question that Carnap considered the epistemological primacy of the autopsychological beyond dispute. This presupposition was not only disputed in the protocol sentence debate, but the notion of epistemological primacy itself was on the line.
2.2 Physicalism
On the next stage, in contrast to the Aufbau, the problem of a confirming base has been in focus from the beginning. This new standpoint is, of course, the physicalistic one.13 Specifically, the thesis of physicalism as presented by Carnap here for the first time divides into two sub-theses: first, the whole of science is capable of being expressed in physicalistic language, that is, the physicalistic language is the comprehensive system-language. This thesis denies that there are areas of science which cannot be framed in physicalistic terms, most notably psychology. Our focus here, however, is on the second sub-thesis: the protocol-language is also part of the one and only system-language. While starting with a two-language conception (protocol-language and system-language), we end up with the all-encompassing physicalistic system-language.
In the present state of research it is not possible to characterize this language with greater precision, i.e. to specify its vocabulary, syntactical forms and rules.15
More specifically, it is an open question as to whether protocol sentences speak of simple sensations, complexes of sensations (âGestaltenâ, or entire sensory fields, or even the total experience of a moment), or material objects.16 While underdetermined with regard to both semantic content and syntactical form, Carnap offers an epistemological characterization. Protocol sentences âcannot be rejectedâ;17 while all other synthetic sentences have to be justified by reference to protocol sentences, the latter themselves âneeding no justificationâ.18
The first thing to note here is a somewhat careless use of epistemological characterizations. Incorrigibility (impossibility of being rejected) is not the same as not requiring justification. This sloppy use of different epistemological concepts is also to be found in Schlick, who mostly speaks of infallibility, but also uses in one sentence the three diverging characterizations finality, indubitability, and incorrigibility.19 However, the main difficulty at this stage seems to consist in maintaining an epistemological privilege (regardless of which one is preferred in the end) and at the same time adhere to the thesis of translatability. To put it in a more formal way, the three propositions
- âProtocol sentences are epistemologically privileged
- âProtocol sentences are translatable into physicalistic system sentences
- âPhysicalistic system-sentences are epistemologically all on a par (fallible, need to be justified etc.)
do not go together.20
2.3 âOn Protocol Sentencesâ
Carnapâs conception of two languages â an epistemologically privileged protocol-language and a physicalistic system-language â united only afterwards, soon came under attack by Neurath, exactly along the lines of the argument just mentioned. According to Neurath, protocol sentences have to be conceived as physicalistic from the beginning. Therefore, an epistemological
Firstly, he directly rejected Neurathâs argument. Surely, one cannot hold on to contradicting sentences. But Neurath is misguided to conclude from this general fact that in the case under discussion at least one protocol sentence has to be abandoned. The whole discussion takes place in the system-language. Dropping at least one protocol sentence is but one possibility, while another is to regard at least one of the conflicting sentences to result from a faulty translation. It is not without a certain irony that Carnap, in defense of his conception, employs a strategy Neurath is well-known for: in case of contradiction, amendments are required; but the place where the amendment is to take place is not fixed yet.
- âProtocol sentences outside the system-language and subsequent translation (Carnapâs view at the previous stage).
- âProtocol sentences within the system-language (therefore physicalistic), distinguished by a special syntactical form (Neurathâs view).
- âProtocol sentences within the system-language (therefore physicalistic), distinguished by the role they play. One and the same sentence can be used as a protocol sentence (when taken as provisional endpoint of justification) or as a system-sentence standing in need of justification (Popperâs view).
Note that there is neither an epistemological, nor a semantical or syntactical characterization shared by all possibilities. An epistemological privilege is obtained only for the first one, which differs in respect to semantic content from its rivals. And only the second one is demarcated by a special syntactical form.22 Which one to choose corresponds to different possible ways the
Carnap himself leaned strongly towards the third, the Popperian variant to which he was introduced by Popper himself in personal discussions during their holidays in the Tyrolian Alps immediately before writing his paper. According to this version, protocol sentences are to be characterized conventionally in two respects. First, as already stated, by preferring this version to its two rivals, and second, by the fact that within this version, the characterization of a sentence as a protocol sentence is always relative to a certain concern. In another context (in the case of doubt, or the need for closer examination, etc.), the same sentence can be taken to stand in need of justification by other protocol sentences.23 A protocol sentence is distinguished by the fact that we decide to stop further justification without saying that we have reached an absolute end. We are satisfied to pursue the task of justification up to this point â at least for the moment. This decision can be revised by changing the former protocol sentence into a questioned hypothesis.
For the time being, Neurath seemed to be satisfied. After all, the legitimacy of his own position was acknowledged. In addition, Neurath adhered to the view that no sentence is immune from being dropped, and that acceptance of each single sentence (whether protocol sentence or not) is always a matter of decision. Following the intense discussion in their letters preceding the joint publication of Neurathâs âProtokollsätzeâ and Carnapâs âÃber Protokollsätzeâ at the end of 1932, the issue was no longer a central topic in the correspondence for a while.
It would have been better if, in accordance with my wish, you had not brought up the matter at that time, when everything is still unclear. Now we have the damage: all people find the opportunity to criticize our, especially your formulations, and partly rightly so. We should rather publish only when things are sufficiently clear to ourselves.24
The remark about premature publication refers to both our essays; but not to those on the general problems of physicalism, but to those on protocol sentences. This problem does not seem sufficiently clear to me (even today). I meant that you should not yet have fully addressed this problem before we ourselves are at least halfway clear about it.25
My view on protocol sentences is not yet clearly formulated anywhere. The important points are, however, partly discussed in an English essay I am currently writing.26
These concessions are not easy to understand. One wonders how physicalism can be stated while omitting the question of how to include protocol sentences (as Carnap himself had stated right from the beginning as one major task). Putting this aside, it is obvious that Carnapâs âOn Protocol Sentencesâ27 was by no means his final word on the matter. To be sure, this is an important paper, e.g., the soon to be called âprinciple of toleranceâ is at work here for the first time. Regarding the issue of protocol sentences, however, it is nothing more than a snapshot of a certain stage in Carnapâs development. This remark seems to be indicative, because, up to now, this paper has often been regarded as Carnapâs main contribution to the debate.
I shall now turn to the second part of this paper and attend to Carnapâs final position outlined in the paper mentioned in the quote just given, namely âTestability and Meaningâ.
3 Paris 1935 and âTestability and Meaningâ
Carnapâs mature position emerged in his talks delivered at the first conference for Unity of Science in Paris in September 1935, which appeared in print the year after.28 These papers mark a significant milestone in his development, e.g. by the adoption of the semantic conception of truth (from now on strictly delimiting truth from confirmation), or the introduction of non-eliminative reduction instead of definability, a decisive step towards a liberalization of empiricism. At the same time he was working on the book-length essay âTestability and Meaningâ, elaborating in detail some of these new ideas.29 Taken together, in these writings, not only did Carnap develop a new account of protocol sentences, but also delivered a re-consideration of the whole discipline formerly called âepistemologyâ. It is to bear in mind that one cannot
Epistemology, up to now, is âan unclear mixture of psychological and logical componentsâ.30 The main target of this critical remark is Schlick (although not named) who stated that the issue of protocol sentences is a psychological problem.31 But Schlick was not the only one who fell prey to this confusion â Carnap confessed that this unclear mixture is also to be found in his own earlier work.32
Most notably, the adoption of an empiricist standpoint is conventional. It is misguided to view empiricism as an assertion. Much better, empiricism is formulated as the proposal to adopt a certain form of language. This restricted form of scientific language requires that descriptive predicates and synthetic sentences are admitted only if there is a certain connection to an experiential base.33 Once this decision is made, one is obliged to state such a base. This task cannot be achieved by pure convention, since empiricism demands for testing by means of experience and not by means of arbitrarily chosen sentences. In other words, as long as protocol sentences are not characterized, empiricism itself does not occupy a distinguished position. In any case, protocol sentences cannot simply be chosen in a completely free way. This is an important self-correction on the part of Carnap because that was exactly the way he went about it in âOn Protocol Sentencesâ. In another respect, Carnap still adhered to his former view, namely that protocol sentences are neither distinguished by a special syntactical form, nor by their semantic content. As before, in the end, it is a matter of convenience, a matter of considerations of usefulness, whether protocol sentences are formulated in a phenomenalistic or physicalistic form. In other words, semantically, protocol sentences are not uniquely determined. Carnap now envisages the following possibilities: protocol sentences must contain observable predicates; these predicates can be:
- a)psychological predicates in phenomenalistic language
- b)psychological predicates in physicalistic language
- c)physicalistic predicates of the âThing-Languageâ
This is close to the three possibilities Carnap envisaged in 1932 (cf. above, sec. 2.3), but there are important differences. The possibility of completely freely-chosen protocol sentences is no longer a real option, instead there is now a
3.1 The Criterion of Quick Decidability
The crucial question now is how to characterize âobservabilityâ, i.e. how to make clear what is meant exactly by âconfrontation with observationâ. In admitting the indispensability of direct testing, Carnap appears to be closer to Schlick, but unlike Schlick, Carnap neither views this process as a comparison between sentence and reality34 nor does he adhere to Schlickâs thesis that this procedure necessarily takes place in the private domain of oneâs own experiences. Instead, to clarify the nature of âconfrontingâ and therefore to obtain a criterion of demarcating protocol sentences, we have to turn towards science: âThe description of that procedure is not a matter of logic but is itself empirically-scientific.â35
This is nothing but a straightforward consequence of Carnapâs conception of philosophy: Everything that remains of philosophy is logic of science. Questions of logic of science are nothing but problems of language. Since confirmation by experience resp. confrontation with reality transcends the limits of language, it should not be supposed to be part of logic of sciences; it rather belongs to science itself and not to philosophy proper. Although Carnap had at the time already outpaced the general restriction to syntactical features of his âsyntactical periodâ (culminating in Carnap 1934/1937), this point can be illustrated by the distinction of the material and the formal mode of speech: Whereas the factual statements belong to the material mode of speech, the proper philosophical statements are in the formal mode, concerning syntactical features of language.
Therefore, protocol sentences cannot be determined by purely philosophical considerations. It is an empirical, scientific task to determine what it is for a sentence to be tested by direct confrontation with experience. This distinctive feature should be delivered by â what I call â the criterion of quick decidability:
We shall speak of âdirectly testable statementâ when circumstances are conceivable in which we confidently consider the statement so strongly confirmed or else disconfirmed on the basis of one or very few observations that we would either accept or reject it outright.36
A predicate âPâ of a language L is called observable for an organism (e.g. a person) N, if, for suitable arguments, e.g. âbâ, N is able under suitable circumstances to come to a decision with the help of few observations about a full sentence, say âP(b)â, i.e. to a confirmation of either âP(b)â or â~P(b)â of such a high degree that he will either accept or reject âP(b)â.37
The whole of epistemology is structured that way: the philosophical part, dealing with relations between sentences, including those sentences which are taken to be basic. This part, of course, is the logic of science. Here, we are solely concerned with relations between sentences; there is no place left for an epistemological subject of whatever kind. In this sense, traditional epistemology is completely abandoned, it is epistemology without a knowing subject.38 On the other hand, within the logic of science there is no possibility of distinguishing protocol sentences. To put it more precisely: there is no possibility of distinguishing basic sentences in a way not wholly conventional, which â as Carnap admits â does no justice to empiricism. To take an empiricist stance, it is indispensable to opt for direct confirmation. Therefore, protocol sentences are to be distinguished within the realm of the empirical by means of scientific investigations.
It is instructive to relate this interplay between science and logic of science with the characterization of epistemology delivered in the Aufbau. There, too, Carnap held that the only epistemological task is to investigate the relations between propositions.39 But now Carnap acknowledges the need for supplementation â at least, if one wants to do justice to the demand of empiricism â since if the proper philosophical task consists only in the investigation of the relations between different sentences, there is nothing to be said about
There is no sharp line between observable and non-observable predicates because a person will be more or less able to decide a certain sentence quickly, i.e. he will be inclined after a certain period of observation to accept the sentence.40
4 Assessing Carnapâs Final Position: Epistemology Dismissed?
Looking back at his development, it is clear that at the beginning the issue of protocol sentences was not the focus of Carnapâs work. He did not start his career as a marked proponent of empiricism; the Aufbau is designed as a neutral position to overcome traditional disputes. Neither the function of providing contact with reality, nor the function of terminating points of confirmation can be bestowed on the (constructional) base of the Aufbau. Protocol sentences, which should exactly serve that purpose, do not seem to be capable of being integrated in this system. Similarly, protocol sentences appear to be a residual problem for the next stage, namely physicalism. The main problem here is to make sense of the idea of endpoints of confirmation needing no
No wonder that Carnap was very sensitive to the contributions of his fellow scholars at the time. He integrated what he took to be insights of his friends and opponents in his mature standpoint, but also disagreed in other points with each of them. Like Schlick, Carnap tried to make sense of the notion of direct confirmation, but he viewed Schlickâs attempt to be a confused mixture of philosophical and psychological considerations. Like Popper and Neurath, he opposed the infallibility Schlick claimed, taking protocol sentences as semantical complex hypotheses. He joined Neurathâs camp against Popper in holding on to the idea that protocol sentences are not to be distinguished purely conventionally â they are observational reports.42 In opposition to Neurath, he sided with Schlick in rejecting a radical coherentist view of justification by opting for a direct confirmation of protocol sentences (see below); due to that characteristic they are able to function as (provisional) regress stoppers.
Nonetheless, the final stage of Carnapâs development is not only synthesizing but also highly original. One aim of the paper was to show the emergence of a solution in response to what he saw as shortcomings of his earlier positions. In the end, he reached a standpoint which combines different elements at different levels:
It is â and according to Carnapâs metaphilosophical stance, it has to be â a matter of science to determinate protocol sentences; therefore, protocol sentences are characterized in a naturalistic way by the criterion of quick decidability.
A matter of convention is, in the first place, the decision for an empiricist language, which is characterized by admitting descriptive sentences only if
This detachment of the characterization of protocol sentences from their syntactic/semantic features is of utmost importance and can be highlighted by a contextualization as well. This detachment not only clarifies the confusion of the two types of bases of the Aufbau, it also cuts the ties to the elementary propositions of Wittgensteinâs Tractatus. There, the existence of such propositions as simple and indivisible units is a precondition of the picture theory: they are characterized semantically; as Wittgenstein puts it, their existence is assured by logic.44 Apart from the contentious issue as to whether verificationism is at least implicitly contained in the Tractatus itself, the elementary propositions were understood in the Circle, above all, in an epistemological sense. From the perspective of further development, it was Feyerabend who made prominent use of Carnapâs ideas. Under the name âpragmatic theory of observationâ this view became a key idea in his crusade against what he took to be the prevailing sterile, ahistorical and dogmatic view.45
4.1 The Criterion of prima facie Credibility
What has been achieved so far is a demarcation of protocol sentences, which fits into Carnapâs mature conception of philosophy. The criterion of quick decidability achieves the distinction of protocol sentences in terms of their formation. Surely it is a scientific task to mark up those sentences that are held
Now, if confirmation is to be feasible at all, this process of referring back to other predicates must terminate at some point. The reduction must finally come to predicates for which we can come to a confirmation directly, i.e. without reference to other predicates.46
A statement established on the basis of the first operation [confrontation with experience, J.F.] is held as (sufficiently strongly) confirmed as long as in the second operation [confrontation with other statements, J.F.] no statements are found which were previously established by confirmation but are incompatible with the statement under consideration.47
It cannot be overstated that in acknowledging the need for direct confirmation, Carnap still holds on to a basic tenet of empiricism in a more traditional sense. This differs sharply from strategies in which (provisional) termination of justification is sought by appealing to further instances. There are different ways such a strategy can take: one might appeal to the reliability of the process by which protocol sentences are acquired, or simply by referring to our practice as codified has proved to be successful. No matter how such accounts are worked out in detail, they all have in common that the âreliabilityâ of protocol sentences results from other factors. Therefore, protocol sentences are not terminating points of justification; there is no terminating point at all since the reliability itself (or our hitherto successful practice) cannot simply be taken for granted. Stating these conditions, we get hypotheses which stand themselves in need of empirical justification, a process that refers to protocol-sentences etc., etc. It is out of the scope of this paper to discuss such âholisticâ, âcoherence-theoreticalâ accounts in more detail, for the present context it is sufficient to state that on any such account, it is denied offhand what Carnap is looking for, i.e. direct confirmation.
Letâs turn to the criterion of prima facie credibility itself. At the least, it seems to be a traditional epistemological principle. In contrast to the criterion of quick decidability, its integration in Carnapâs metaphilosophical stance has to
Slightly simplified, the criterion of prima facie credibility states that protocol sentences are confirmed as long as there are no contradicting sentences. That cannot mean that protocol sentences are confirmed by other sentences. Therefore, they have to be taken as points where there is â at the moment â neither negative nor positive evidence; protocol sentences are simply unexamined, âneutralâ items with nothing for or against them. But surely this neutrality does not suffice for confirmation, otherwise any suitable, purely fictional hypothesis would count as prima facie justified. The latter case is typical for the demand of confirmation, and it is odd to demand for confirmation in this case and not in the case of neutral protocol sentences.
Nor is it of any help to emphasize the provisional nature of this purported confirmation. The phrase âas long as there are no contradicting sentencesâ stands in need of clarification. For each contingent sentence, a contradicting sentence can easily be âfoundâ, most simply by negation. Hence, is it meant that those contradicting sentences are in fact entertained? Just holding a refuting sentence is still not sufficient, otherwise any sentence contradicting a protocol sentence would count as rebutting instance, however futile it may be, if it is hold to be true by â let us say â a mentally deranged person. To refute a protocol sentence, we demand that the contradicting sentence has positive evidence in favor of it; Carnap acknowledges this by admitting that the refuting evidence must consist of sentences âwhich were previously established by confirmationâ. But what he does not seem to see is that his explication of empirical confirmation runs in a circle. In brief, absence of contradicting evidence will not do, positive evidence is required. This condensed skeptical argumentation49 may seem unfair to Carnap, treating him as a proponent of fallible foundationalism, to use contemporary terminology. But this problem is a straightforward outcome if one takes the relation of confirmation as having a direction, like Carnap himself â insofar the accusation of circularity is a serious one. Furthermore, it should be noted that the situation is not defused by the fact that, for Carnap, opting for an empiricist language is a conventional choice. The issue here is raised within that empiricist framework and does not affect the framework itself.
It is out of the question that Carnap did not want to get tangled up in such traditional epistemological disputes. His new conception of philosophy is, at
Obviously, we have reached one of the big issues in modern epistemology: the difficulty of placing the criterion of prima facie credibility reflects the problem of the nature of epistemological principles. Due to their normativity, they do not seem to be reducible, neither to conventional rules of language (at least, if one wants to stick to the idea of empirical confirmation), nor to science. My aim here was to state that there is more than just a little residual problem of minor importance; that getting rid of traditional epistemological disputes is not so easy. By no means, however, do I want to say that this problem constitutes a definite refutation of Carnap. Some scholars look with optimism to the prospect of incorporating Carnap within the naturalistâs camp.50 In any case, with his mature view on protocol sentences, Carnap stands right at the beginning of this debate, albeit, in several respects, he seems to be caught in the middle.
E.g., Neurath bans any talk of reality as metaphysical; Popperâs falsificationism does not acknowledge any positive value of confirmation.
Carnap 1928/2003; hereafter simply Aufbau.
A possible exception is Hempel who stated that âthere is no essential difference left between protocol statements and other statementsâ (1935/2000, p. 19). It is only a possible exception since it is not entirely clear if Hempel should be read as a proponent of a conventionalist characterization (see below). On Hempelâs proposal within the context of the debate, see Uebel 2007, sec. 9.3.1.
Aufbau, § 58.
Aufbau, § 75, 78.
Aufbau, § 65.
This difference is stated very clearly in Russell 1924/2007, p. 325f.
Aufbau, § 177.
Aufbau, § 58.
Aufbau, § 178.
Cf. Friedman 1992/1999, and Richardson 1998, especially Chap. 8.
For a more detailed account on the tension between basic tenets of the Aufbau and verificationism cf. Friedl 2021.
For the shift from the Aufbauâs structural approach to physicalism, see Sauer 1989.
Carnap 1931/1934, pp. 42â45.
Carnap 1931/1934, p. 45.
Carnap 1931/1934, p. 46f.
Carnap 1932/1959, p. 19.
Carnap 1931/34, p. 45.
Schlick 1935b/1979, p. 412; for discussion see Friedl 2013, Chap. v, Sec. 3. The first exposition of the theory of âKonstatierungenâ (âaffirmationsâ) is Schlick 1934/1979.
Cf. Coffa 1990, p. 358.
Neurath 1932/1983, p. 95.
Put more precisely, Carnap as well as Popper demand protocol sentences to be concrete sentences resp. singular existential sentences. Albeit this is a syntactical characterization, protocol sentences are, of course, not singled out by this requirement.
According to Popper, conventionalism is the thesis that laws of nature (universal propositions) are fixed by convention, whereas he himself treats basic sentences (singular propositions) that way (1935/2002, p. 91f.). But one wonders how this differentiation is supposed to work: any universal proposition can be defended, simply by choosing the appropriate basic sentences; therefore, conventionalism in regard to basic sentences entails conventionalism in regard to laws; cf. Stegmüller 1969, pp. 360â62.
âEs wäre besser gewesen, Du hättest, meinem Wunsch entsprechend, damals nicht schon die Sache aufgerollt, wo doch alles noch unklar ist. Jetzt haben wir den Schaden: alle Leute finden Gelegenheit, unsere, besonders Deine Formulierungen zu kritisieren, und zum Teil mit Recht. Wir sollten lieber erst dann veröffentlichen, wenn uns selbst die Dinge hinreichend klar sindâ (Carnap to Neurath, May 18, 1934, asp-rc-029-10-63).
âDie Bemerkung von vorzeitiger Veröffentlichung bezieht sich auf unser beider Aufsätze; aber nicht auf die über die allg. Probleme des Physikalismus, sondern auf die über Protokollsätze. Dieses Problem scheint mir (sogar heute noch) nicht hinreichend klar. Dieses Problem, meinte ich, hättest Du noch nicht in vollem Umfange aufrollen sollen, bevor wir selbst uns wenigstens halbwegs klar darin sindâ (Carnap to Neurath, June 8, 1934, asp-rc 029-10-61).
âMeine Ansicht über Prot.-Sätze ist allerdings noch nirgends deutlich formuliert. Die wichtigen Punkte werden aber zum Teil in einem engl. Aufsatz besprochen, den ich gerade schreibeâ (Carnap to Neurath, June 22, 1935, asp-rc 029-09-44).
Carnap 1936a, 1936b, 1936c/1949.
Carnap wrote the first version in spring 1935, the final version in spring 1936 (Carnap to Olga Neurath, August 27, 1936, asp-rc 102-52-33).
â[E]ine unklare Mischung aus psychologischen und logischen Bestandteilenâ (Carnap 1936a, p. 36).
Schlick 1935a/1979, p. 404; see also Schlick to Carnap, April 16, 1935 (asp-rc 102-70-14).
Carnap 1936a, p. 36.
Carnap 1936/37, p. 33.
For Carnapâs critique of Schlickâs conception of comparison, see Carnap 1936c/1949, p. 125f.
Carnap 1936c/1949, p. 124.
Carnap 1936c/1949, p. 124.
Carnap 1936/37, p. 454f.
See Uebel 2018, p. 372; Uebelâs account of the transformation of Carnapâs epistemology has been a valuable suggestion for this paper; a point of disagreement is treated within the concluding section.
Cf. Aufbau, § 178 (quoted above, sec. 2.1).
Carnap 1936/37, p. 455. Carnap does not mention that the criterion cannot function as a genuine definition of âobservableâ if only because of the circularity contained in it.
See above, sec. 2.3.
In addition, Carnap raised the objection against Neurath that protocol sentences have to deal with observations themselves and are not constructed as sentences on other sentences. Neurath took Carnapâs charge as a misunderstanding; for a reading of Neurathâs somewhat strange, interlaced protocol sentences according to which Carnapâs charge is really a misunderstanding, see Bentleyâs chapter in this book.
Carnap 1936/37, p. 10f. Since Carnap still allows such purely subjective languages as starting points (while now admitting that such a choice is inconvenient), it is misleading to view his abandonment of âmethodological solipsismâ as the decisive step beyond the Aufbau. After all, it was only due to misunderstandings that he finally dropped that term; cf. Carnap 1936/1937, p. 423f.
Wittgenstein 1921/1961, 5.5562; for a summarizing account of the Tractatusâ doctrine of elementary propositions cf. Glock 1996, pp. 102â107.
Most prominently in Feyerabend 1962/1981; for discussion of the relation to Carnapâs theory of protocol sentences, see Oberdan 1990 and Kuby 2018.
Carnap 1936/37, p. 456.
Carnap 1936c/1949, p. 125.
By identifying this criterion as a necessary supplement, I differ from Richardson (1998, p. 216) and Uebel (2018, p. 371) who view epistemological issues of whatever kind to be already done away by the naturalistic and conventionalist elements of Carnapâs final position.
The argumentation is essentially a summary of Rutte 2000, pp. 89â91, who raises these objections against Popper and his school.
Uebel 2018, p. 375f.; as already noted, Uebel does not pin this down on the criterion of prima facie credibility.
Acknowledgments
This work was supported by the Austrian Science Fund (fwf): P 30377. I am grateful to all participants of the conference for their comments on the talk on which this paper is based, and especially to Thomas Uebel for extensive critical comments on an earlier version.
asp-rc = Archives of Scientific Philosophy, Rudolf Carnap Papers, University of Pittsburgh.
Carnap, R. (1928/2003) Der logische Aufbau der Welt, translated as The Logical Structure of the World, Open Court: Chicago/La Salle, Ill.
Carnap, R. (1931/1934) âDie physikalische Sprache als Universalsprache der Wissenschaftâ, translated as The Unity of Science, London: Kegan Paul, Trench, Trubner & Co.
Carnap, R. (1932/1959) âPsychologie in physikalischer Spracheâ, translated as âPsychology in Physical Languageâ, in Ayer, A.J. (ed.), Logical Positivism, New York etc.: The Free Press, pp. 165â198.
Carnap, R. (1932/1987) âÃber Protokollsätzeâ, translated as âOn Protocol Sentencesâ, Nous 21, pp. 457â470.
Carnap, R. (1934/1937) Logische Syntax der Sprache, rev. ed. translated as The Logical Syntax of Language, repr. Chicago: Open Court 2002.
Carnap, R. (1936a) âVon der Erkenntnistheorie zur Wissenschaftslogikâ, in Actes du Congrès International de Philosophie Scientifique, Sorbonne, Paris 1935, fasc. I: Philosophie Scientifique et Empirisme Logique, Paris: Hermann & Cie, pp. 36â41.
Carnap, R. (1936b) âÃber die Einheitssprache der Wissenschaft. Logische Bemerkungen zum Projekt einer Enzyklopädieâ, in Actes du Congrès International de Philosophie Scientifique, Sorbonne, Paris 1935, fasc. ii: Unité de la Science, Paris: Hermann & Cie, pp. 60â70.
Carnap, R. (1936c/1949) âWahrheit und Bewährungâ, translated with addenda as âTruth and Confirmationâ in Feigl, H. and Sellars, W. (eds.), Readings in Philosophical Analysis, New York: Appleton-Century-Crofts, pp. 119â127.
Carnap, R. (1936/37) âTestability and Meaningâ, Philosophy of Science 3, 1936, pp. 419â471; 4, 1937, pp. 1â40.
Coffa, A. (1990) The Semantic Tradition from Kant to Carnap. To the Vienna Station, edited by L. Wessels, Cambridge: Cambridge University Press.
Feyerabend, P. (1962/1981) âExplanation, Reduction and Empiricismâ, reprinted in Feyerabend, Realism, Rationalism and Scientific Method, Cambridge: Cambridge University Press, pp. 44â96.
Friedl, J. (2013) Konsequenter Empirismus. Die Entwicklung von Moritz Schlicks Erkenntnistheorie im Wiener Kreis, Vienna/New York: Springer.
Friedl, J. (2021) âRealism and Anti-Realism in Young Carnapâ, in Damböck, Ch. and Wolters, G. (eds.), Der junge Carnap in historischem Kontext: 1918â1935/Young Carnap in an Historical Context: 1918â1935, Cham: Springer, pp. 149â168.
Friedman, M. (1992/1999) âEpistemology in the Aufbauâ, reprinted in Friedman, Reconsidering Logical Positivism, Cambridge: Cambridge University Press, pp. 114â162.
Glock, H.J. (1996) A Wittgenstein Dictionary, Oxford: Blackwell.
Hempel, C.G. (1935/2000) âOn the Logical Positivistsâ Theory of Truthâ, reprinted in Hempel, Selected Philosophical Essays, edited by R. Jeffrey, Cambridge: Cambridge University Press, pp. 9â20.
Kuby, D. (2018) âCarnap, Feyerabend, and the Pragmatic Theory of Observationâ, hopos: The Journal of the International Society for the History of Philosophy of Science 8, pp. 432â470.
Neurath, O. (1932/1983) âProtokollsätzeâ, translated as âProtocol Statementsâ, in Neurath, Philosophical Papers 1913â1946, edited by R.S. Cohen and M. Neurath, Dordrecht: Reidel, pp. 91â99.
Oberdan, T. (1990) âPositivism and the Pragmatic Theory of Observationâ, psa: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1990, vol. 1, pp. 25â37.
Popper, K. (1935/2002) Logik der Forschung, trans. with addenda as The Logic of Scientific Discovery, London/New York: Routledge.
Richardson, A.W. (1998) Carnapâs Construction of the World. The Aufbau and the Emergence of Logical Empiricism, Cambridge: Cambridge University Press.
Russell, B. (1924/2007) âLogical Atomismâ, reprinted in Russell, Logic and Knowledge .Essays 1901â1950, edited by R.C. Marsh, Nottingham: Spokesman, pp. 323â343.
Rutte, H. (2000) âEinige erkenntnistheoretische Fragen an den kritischen Rationalismusâ, Geschichte und Gegenwart 19, pp. 79â99.
Sauer, W. (1989) âCarnap 1928â1932ââ in Gombocz, W.L., Rutte, H. and Sauer, W. (eds.), Traditionen und Perspektiven der analytischen Philosophie. Festschrift für Rudolf Haller, Vienna: Hölder-Pichler-Tempsky, pp. 173â186.
Schlick, M. (1934/1979) âÃber das Fundament der Erkenntnisâ, translated as âOn the Foundation of Knowledgeâ, in Schlick (1979), pp. 370â387.
Schlick, M. (1935a/1979) âFacts and Propositionsâ, reprinted in Schlick (1979), pp. 400â404.
Schlick, M. (1935b/1979) âÃber âKonstatierungenââ, translated as âOn âaffirmationsââ, in Schlick (1979), pp. 407â413.
Schlick, M. (1979) Philosophical Papers, vol. ii, edited by H. Mulder and B. van de Velde-Schlick, Dordrecht etc.: Reidel.
Stegmüller, W. (1969) Metaphysik, Skepsis, Wissenschaft, 2nd ed., Berlin etc.: Springer.
Uebel, T. (2007) Empiricism at the Crossroads: The Vienna Circleâs Protocol-Sentence Debate, Chicago/LaSalle, Ill.: Open Court.
Uebel, T. (2018) âCarnapâs Transformation of Epistemology and the Development of His Metaphilosophyâ, The Monist 101, pp. 367â387.
Wittgenstein, L. (1921/1961) Tractatus Logico-Philosophicus, translated by D.F. Pears and B.F. McGuinness, New York: Humanities Press.