1 Protocol Statements and the Bipartite Metatheory Conception
What Uebel has called the bipartite metatheory conception attributes to Neurath and Carnap ‘a broadly shared program for a bipartite metatheory serving as a joint replacement for philosophy’.2 This scientific metatheory is a second-order inquiry into science itself, composed of two methodologically distinct parts; the empirical study of science, including the history, psychology and sociology of science, and the formal study of the logical structures of axiomatized scientific theories and terminology. These two meta-theoretical
Protocol sentences were understood by the members of the Vienna Circle as the basic evidence statements of science, those used in testing and confirmation. Though frequently interpreted as incorrigible statements of immediate experience (as utilised by traditional foundationalist empiricism) by both the Vienna Circle’s contemporaries and modern commentators, there was no consensus within the Vienna Circle as to how protocol sentences ought to be understood. Rather, the protocol sentence debates of the early '30s involved extensive discussion of the form, status and purpose of protocol sentences within science. And whilst this may appear on the surface as a narrow technical debate, really under discussion were competing conceptions of scientific knowledge and the process of justification. Often (although not always) implicit in the debates over the correct understanding of protocol sentences was a deeper disagreement about the nature of scientific knowledge, and the appropriate methodology for philosophy. Discussions of these competing conceptions of protocol statements have typically (and with good reason) focused on the crucial period of the debates, between 1932 and 1934. When specifically considering the debate between Neurath and Carnap over protocol sentences, one’s attention is immediately drawn to the interaction of 1932, particularly Neurath’s Protocol Sentences and Carnap’s reply On Protocol Sentences. Prima facie, this seems like the crescendo of the Neurath-Carnap debate; after 1932, it re-orients to a clash between Neurath and Schlick. And from 1935 onwards, the disagreements and discussions between Carnap and Neurath centre on Carnap’s embrace of Tarskian semantics. The issue of protocol sentences may therefore appear to have lost its importance in Neurath and Carnap’s thinking.
Otto’s protocol at 3:17 o’clock: [Otto’s speech-thinking at 3: 16 o’clock was: (at 3:15 o’clock there was a table in the room perceived by Otto)].5
But he later adopts a quadruple-layered structure; ‘Karl’s protocol: Karl formulates: Karl is seeing: In the room there is a round table’.6 The advantage of this change, which will be made clearer below, is that it separates out the act of observation from the object under observation. It is this quadruple-embedded conception that I will refer to as Neurath’s mature conception. By contrast, Carnap’s notion of a protocol sentence undergoes frequent and significant alterations, especially during the debate of the early '30s. Over the course of the debate, Carnap revises the form, role and epistemic status of protocol statements. I will therefore start with a brief overview of Carnap’s developing
2 Carnap’s Changing Conception
For current purposes, three main periods in the development of Carnap’s conception of protocol statements can be identified.7
- 1.Early-Carnap: pre-1930
- 2.Middle-Carnap: 1930-35
- 3.Mature-Carnap: 1936 onwards
The early Carnap includes the Aufbau era, and predates the protocol sentence debates. The middle is the period of intense fluctuation of his conception, with almost every different work published in this period involving a slight change in either his ideas or the deployment of terminology. This includes the crucial period of the protocol sentence debates. The mature period begins with Testability and Meaning in 1936, where Carnap develops a conception which is essentially retained from then on, as other issues took precedence over protocol sentences.8
Prior to the protocol sentence debates, in the Aufbau Carnap utilised sentences in the auto-psychological language describing the immediate given (at bottom remembrances of similarity relations), although he recognises the possibility of constructing a similar system on a physicalist basis.9 In 1930, Carnap still allows both methodological materialism and solipsism, treating physicalist and auto-psychological languages as equally basic, although he continues to refer to statements as ‘reducible to the given’.10 The auto-psychological protocol statements were understood as incorrigible. However, translation into the physicalist language was necessary for scientific purposes, because only a physicalist language ‘makes inter-subjective knowledge possible’.11 Carnap
By early 1932 in Unity of Science, Carnap concedes the inter-translatability of the physical and auto-psychological languages, discarding the translational primacy of the auto-psychological. The protocol language was no longer understood as a private phenomenological language, but as a sub-language of the physical universal language.14 But statements still required translation into the auto-psychological language for epistemological purposes, primarily justification. And Carnap still maintains the conception of “the given”, describing protocol statements as a ‘direct record of a scientist’s … experience’.15 Auto-psychological protocol statements, despite losing their privacy, retain their epistemic privilege; ‘a protocol sentence, being an epistemological point of departure, cannot be rejected’.16 Carnap still allows for primitive protocol statements, assuming a ‘sharp (theoretical) distinction between the raw material of scientific investigation and its organization’.17 The auto-psychological protocol language therefore is still somewhat distinct from the physical language, despite Carnap’s claims that it is simply a sub-language.
The major change in Carnap’s conception of protocol statements, in which these prior commitments are abandoned, came about in late 1932 in On Protocol statements. Here Carnap finally withdraws the epistemological primacy and privilege of the auto-psychological; physicalistic statements can now serve epistemological purposes. With this epistemological privilege revoked, protocol statements also lost their certainty and incorrigibility. Now the form of protocol statements is completely unrestricted; ‘it is a matter of decision which sentences one wants to use at various times … as protocol sentences’.18
The late 1932 conception is partially an overreaction to Neurath’s anti-absolutist, anti-foundationalist arguments for the fallibility and revisability of protocol statements, captured by what Haller called the Neurath principle.21 But Carnap was also influenced by Popper who, he claimed, took this anti-absolutism a step further than Neurath had.22 In so doing, Carnap also takes conventionalism a step too far. Carnap sees in Popper’s rejection of a ‘last sentence’ the most anti-absolutist form of protocol statement, even more so than Neurath’s.23 But we should remember that Neurath only has “last sentences” in a very mitigated sense; his protocol statements have only a special methodological role in testing, but no distinguished or unique epistemic status. This provides them significance via a particular role in theory testing, one which they are deprived of if we allow any sentence to qualify as a protocol statement. As Neurath shows, one can reject the epistemological privilege Carnap had previously attributed to protocol statements without falling into the total conventionalism of On Protocol statements. It is still a matter of decision as to what qualifies as a protocol, but it is not an arbitrary decision; it is a decision guided by the practical requirements of scientific practice. Whether for these reasons or not, Carnap arrived at a similar conclusion; this completely conventional conception of protocol statements appears only in On Protocol statements.24 But whilst Carnap dials back the conventionalism, his rejection of epistemological privilege is definitive.
3 Mature Positions: Apparent Disagreement
This simplified sketch of the protocol sentence debates, at least the Carnap-Neurath axis, sees Carnap conceding a certain amount to Neurath; abandoning foundationalism, methodological solipsism and phenomenalistic auto-psychological protocol statements as practically impossible, and consequently adopting a physicalist protocol language. The fundamental disagreement about the need for a phenomenal protocol language seems to be overcome, and a broad agreement reached; protocol statements are physicalistic reports, open to revision and accepted by decision. But crucially, Carnap never adopts the specifics of Neurath’s account. Two key elements of Neurath’s proposal are never embraced by Carnap. Firstly, the multiply-embedded bracket structure, and secondly, the requirement of contextualising information contained within the protocol. These requirements are intimately intertwined for Neurath; the purpose of the structure he advocates is to exemplify the contextualisation, as will be demonstrated below. But Carnap rejects both; the multiply-embedded structure was rejected as impractical, and the restrictions Carnap placed on contextual information never required containment within the protocol itself.
As far as your formulations are concerned (which unfortunately are not in agreement with your opinions) I have told you since many years that I cannot accept them and hence I agree with the criticism of these formulations by R[ussell]., Schlick, and many others. In distinction to R., I know your actual conception from conversations; and I am in agreement with it … I agree with R. in his criticism of your triple-involved form of protocol sentences.28
You always tell me, you agree with Russell’s and Schlick’s remarks on my protocol statements, but my statements intended are different. Please, tell me first what you think how I should express my statements properly and then please tell me, why even then they are not in harmony with your opinion. I think it is important to come to some clearness.29
This exchange suggests a lingering uncertainty about protocol statements that is sadly left unresolved, the possibility of further discussion cut short by the fractiousness of their later correspondence and Neurath’s unexpected death. Despite claiming that he is in agreement with Neurath, Carnap goes on to reject his proposal for the form of protocol statements. The question then lingers: did they agree or not? Making sense of this interaction is essential for understanding the compatibility of their mature conceptions of protocol sentences. In the rest of this article, I will explore the nature of this apparent disagreement, and its significance for the bipartite metatheory conception. There are two potential disagreements between Carnap and Neurath, which broadly correspond to the two elements of Neurath’s proposal that Carnap doesn’t adopt. The first is a disagreement about structure and logical form. The second is the issue of what information needs to be contained within the protocol statement itself. The issue of logical form will be discussed first.
4
Thus according to Neurath the data of empirical science are all of the following form: “A certain person (who happens to be myself, but this, we are told, is irrelevant) is aware at a certain time that a little while ago he believed a phrase which asserted that a little while before that he had seen a table.” That is to say, all empirical knowledge is based upon recollections of words used on former occasions. Why recollections should be preferred to perceptions, and why no recollections should be admitted except of thought-words, is not explained.30
Russell’s criticism specifically targets Neurath’s multiply-embedded bracket structure. If Carnap agrees with Russell, as he claims in his letter, then his problem must also be with the structural and formal-logical aspect of Neurath’s proposal. As Carnap admits, he agrees with what he understands as Neurath’s conception of protocol statements (physicalistic, revisable, unprivileged). But this conception, Carnap claims, is in conflict with or is undermined by Neurath’s proposed bracketed structure. This is not the first time Carnap questioned the logical structure of Neurath’s protocol statements. Carnap had expressed similar doubts during the protocol sentence debates, where he criticised Neurath for the impracticality of a protocol sentence with ‘three nested components’.31 He also claims that Neurath’s proposal ‘has the defect, from the point of view of syntax, that a sentence which refers to another contains the other as a clause’.32 Exactly why this troubles Carnap is unclear and unexplored, but again it is rooted in the logical and structural elements of Neurath’s proposal. It has to be conceded that the logic of Neurath’s protocol statements is not made sufficiently clear by Neurath himself. It can also be conceded that Neurath was not always sufficiently explicit about the motivations for the protocol structure he adopts. Carnap’s confusion at the time of their correspondence is understandable. But Neurath would later attempt to make explicit his
Neurath’s pictures should prompt those looking at them to go back and forth between at least two constellations (normally more than two) of elements, figuring out for themselves what the comparison is all about.35
What the comparison with isotype shows is that instantaneous, prima facie understanding was not the expectation. The information to be conveyed is complex, and as such requires a degree of complexity in presentation to fully
What is essential to understand is that Neurath provides a protocol sentence schema, and the structure is designed with a specific schematic purpose in mind. For Neurath, protocol statements are not about truth-conditions, but acceptance conditions. So his schema prioritises the presentation of the factors relevant to the acceptance procedure in a way that maximises both ease of understanding and practical utility. Maximum clarity about when these conditions are met and the ease with which these statements can be put to use in practice are of much greater significance for his purposes than providing his proposals in proper logical form. Consequently, rather than a protocol sentence as a statement of standard predicate calculus, Neurath gives a synoptic schema that visualises the factors and process of protocol acceptance with maximum perspicuity. As opposed to grammatical usage, punctuation is deployed for purposes of visual communication. The price of this maximisation of epistemological perspicuity is a loss of typical logical form. But Neurath’s emphasis on the pragmatics of science over the logic of science justifies this prioritisation.
‘Otto's speech-thinking was at 3:16 o'clock: (at 3: 15 o'clock there was a table in the room perceived by Otto)’ and further: ‘At 3:15 o'clock there was a table in the room perceived by Otto.’ 36
Russell interprets Neurath’s proposal as a conventional sentence reporting an experience, albeit one containing a lot of brackets. Consequently, he expands the brackets in the typical grammatical way. But when we recognise that Neurath’s protocol statements are not the regular sentences Russell treats them as, it becomes clear that the brackets cannot and should not be expanded this way. The brackets are not being used to separate linguistic clauses, but to visually separate distinguishable factors in the appraisal of observation reports.
The most important demonstration of the role Neurath intends his schema to play is provided in his 1941 paper Universal Jargon and Terminology. Part of the paper is a direct response to Russell’s book, published the same year.37 It therefore seems very plausible that Neurath provided this example specifically to dispel the confusion shown by Russell, and by extension Carnap. Even if this was not the case, his examples can play that role for us. Neurath begins by separating ‘the four ‘parts’ of our protocol statement: A (protocol), B (word-thinking), C (zebra), D (person perceiving)’.38 Neurath then gives three example charts for the categorisation of protocol statements:39
A, B, C, D, accepted B, C, D, accepted C, D, accepted D, accepted ‘factual statement’ |
A, B, C, D, accepted B, C, D, rejected C, D, accepted D, rejected ‘type of lying’ |
A, B, C, D, accepted B, C, D, accepted C, D, rejected D, accepted ‘hallucinatory statement’ |
What these charts show is that the multiply-embedded structure of Neurath’s protocol statement is intended to simultaneously exhibit four sub-statements, each of which is essential for the reception of observation reports. These four sub-statements of Neurath’s schema are:40
- (i)protocol (thought [stimulation state {observable fact}])
- (ii)thought [stimulation state {observable fact}]
- (iii)stimulation state {observable fact}
- (iv)observable fact
Following Uebel’s terminology, the sub-statement contained in the inner-most bracket, the statement expressing an observable fact, is referred to as the
Each of sub-statements (i) – (iv) corresponds to one of four conditions for protocol acceptance, as previously explicated by Uebel.42 For an object sentence X, these conditions are:43
- *(i)the institutional condition: somebody made the explicit claim that somebody thought that somebody was stimulated as if she perceived X
- *(ii)the intentional condition: somebody conceptualised that somebody was stimulated as if she perceived X
- *(iii)the sensory condition: somebody was stimulated as if she perceived X
- *(iv)the negative coherence condition: there is no evidence available that contradicts X
Each condition captures a necessary feature of acceptable scientific testimony. Condition *(i) guarantees that the report was entered into the scientific record; reliable experience reports must come from the observers themselves, specifically entering their experiences into the record. Condition *(ii) guarantees the preservation of the intent and content of the report through the retention of the original linguistic and conceptual framework in which the report was couched. This prohibits attempts to translate or re-interpret past reports, in so doing corrupting them.44 Condition *(iii) guarantees speaker sincerity, ruling out lies. Condition *(iv) is a negative coherence condition establishing speaker competency, ruling out hallucinations and errors by guaranteeing an absence of known defeaters. It shouldn’t be understood as a factive condition, as this would be redundant; why bother with the four conditions if we already know the object sentence is true? Rather, it embodies the possibility that reports can be rejected on the basis of background knowledge and theory.
For Neurath the assessment and categorization of protocol sentences is not simply a matter of the acceptability of each bracketed clause, but the
Importantly, the structure Neurath proposes was never intended as a final account of the form of protocol statements. Whilst protocol statements must be physicalistic, are always revisable, and must contain relevant contextualising information, the specific structure Neurath proposes is contingent on how successfully it performs the functions described above and below. Although I will argue for the success of Neurath’s structure, it is quite possible that an alternative structure could display the same information in a more appealing or efficient way. So long as it fulfils Neurath’s physicalistic and practical requirements, Neurath should welcome such a structure as an improvement upon his own proposal.
5 Motivation for Neurath’s Schema
Karl’s protocol: Karl formulates: Karl sees: In the room is a round table.
‘Karl protocolises that Karl formulates that Karl sees that in the room is a round table’ ∧ ‘Karl formulates that Karl sees that in the room is a round table’ ∧ ‘Karl sees that in the room is a round table’ ∧ ‘In the room is a round table’
Obviously this sentence is incredibly unwieldy.46 But more importantly, the additional complexity masks the inter-relations of the four sub-statements. When rendered explicit like this, we can no longer see the woods for the trees. Such a logical form would be so complex that our ability to grasp the connection between parts and whole is lost. The desired synoptic understanding is not achieved. By contrast, Neurath’s schema exhibits the whole and the parts simultaneously, thereby highlighting the inter-connections, and allowing easier engagement by the observer.
This [protocol] sentence is so constructed that, after ‘deletion of brackets’, further factual sentences appear, which, however, are not protocol sentences: ‘Otto’s speech-thinking was at 3:16 o’clock: (at 3:15 o’clock there was a table in the room perceived by Otto)’, and further: ‘At 3:15 o’clock there was a table in the room perceived by Otto’.47
Karl’s protocol: Karl formulates: Karl sees: In the room is a round table.
Karl formulates: Karl sees: In the room is a round table.
Karl sees: In the room is a round table.
In the room is a round table.
This allows us to check *(iv). These successive deletions allow all four conditions to be quickly and easily checked, allowing for simple categorisation of protocol statements, via charts like those given above. In real-world situations, deletion could be achieved by covering the “deleted” clauses with your hand, or striking them through with a pen removal. This process could be streamlined and visualised even further. Taking Neurath’s charts from above, we could supplement “Accepted” with a tick and “Rejected” with a cross. During the process of successive deletions, one could simply leave a tick or cross for each successive check. The result could then be immediately compared with the categorisation charts. This demonstrates how Neurath intended these protocol statements to be made use of practically and visually, in a way that sentences in proper logical form cannot. Most importantly for our purposes though, in
6 The Methodology and Logic of Science
And now I ask, how we may speak of accepting something as a “lie”, you know I answer, when I accept X says this is brown, when I accept X says internally this is black and not brown, we may accept tentatively X is a liar. If we accept X is saying I see [a] brown table and X is internally saying I see [a] brown table, and we do not accept the statement here is a brown table (as a combined statement, as it were) then we call the X-statement a dream statement or an illusion statement.49
I am in complete agreement with your description of the scientific procedure. I should classify this as belonging to the methodology of science.50
This is a confirmation that Carnap accepts Neurath’s account of the reception and categorisation of observation reports. This should not come as a surprise however, as Carnap repeatedly emphasised the importance of the pragmatics
7 The Processing of Observation Reports
If we incorporate part of the above mentioned protocol, the statement “in the room was a table perceived by Charles” along with the whole protocol into the body of science, then we can speak of a ‘reality formulation’, whereas we would speak of a ‘dream or hallucination formulation’, if we accept the whole protocol but not the part “in the room was a table perceived by Charles”.53
Karl’s protocol: Karl formulates: Karl is seeing: In the room there is a round table.
For the sake of argument we will assume it is accredited. As an accredited protocol, the whole statement can be added to the protocol bank. For the sake of argument, let us also accept that this protocol is valid. In this case, as well as adding the whole sentence to the protocol bank, we are licenced to extract
Karl’s protocol: Karl formulates: Karl sees: In the room is a round table.
and
Karl’s protocol: Karl formulates: Karl touches: In the room is a round table.
are both statements featuring the part:
In the room [is] a round table.56
This passage makes clear that the object sentence embedded within a protocol is a detachable component of it. In Neurath’s example, we have two different protocol sentences with the same object sentence. Both supply the same piece of scientific data (about the table) but the reports supplying the data are different, one being prompted by a visual stimulus and the other by a tactile stimulus.57 Clearly the object sentence itself can be meaningfully separated from the
Crucially, the resulting datum is a simple physicalist statement about medium-sized objects, expressing a spatio-temporal state of affairs. The complexly structured multiply-embedded schema is a means for delivering this simple physicalist datum. Contrary to Russell’s misinterpretation then, Neurath’s account does not render the data of empirical science implausibly complex. It is the protocol schema, and the process of acceptance it embodies, that exhibits such complexity. This complexity is not Neurath’s addition, but simply a reflection of the multiplicity of factors relevant to deciding on whether or not reports should be accepted. But the complexity of the delivery mechanism does not undermine the simplicity of the datum it supplies. It is this important clarification that I think Carnap overlooks, and which I think explains his disagreement. Carnap, like Russell, understood Neurath’s protocol statements to add unnecessary complexity to what should be simple physicalistic statements. We can now see that he was mistaken.
8 Difference of Terminology
With an understanding of Neurath’s account of acceptance in place, we can finally highlight a subtle but significant difference in Carnap and Neurath’s usage of “protocol”: Carnap’s protocol statements, physicalist statements about observable mid-size objects are the object sentences of Neurath’s protocol statements. Carnap gives the example protocol statement ‘a black round table’.58 This is strikingly similar to the object sentence of Neurath’s example ‘Karl’s protocol: Karl formulates: Karl sees: In the room is a round table’.59 Is this just a coincidence? Or is this different usage indicative of a conceptual disagreement? No on both accounts. But to see why, it is helpful to frame the issue differently. According to the bipartite metatheory interpretation, scientific metatheory is composed of two parts; the pragmatics of science as practiced by Neurath and the logic of science as practiced by Carnap. So which part of the metatheory do protocol statements fall under? According to Carnap’s usage, they belong to the logic of science. But according to Neurath, they belong to the pragmatics of science. Again we have the appearance of theoretical disagreement, but really this difference is only terminological.
As to the question of what a protocol statement “really” is, neither Carnap or Neurath would have entertained such a debate. Both recognised the protocol debate as one of competing proposals. Now we know their proposals are not theoretically incompatible, but may require a change in terminology from one party. And I think for the purposes of terminological clarity, Neurath’s proposal ought to be adopted. Firstly, Neurath’s proposal has brought about the Neurathian terminology utilised throughout the discussion above. But perhaps more importantly, Neurath’s usage of “protocol sentence” adheres more closely to the original intention; those basic evidence statements of science. As we saw, significant processing is required to arrive at Carnap’s protocol sentences. The need now is for a term to replace Carnap’s use of protocol. A tentative proposal of a name for evidence statements as utilised within the logic of science is “data sentences”.
9
We can now finally return to Carnap’s rejection of Neurath’s requirement that protocol statements contain the contextual information to answer the question ‘When, where, and how?’60 Carnap specifically rejects Neurath’s requirement for ‘designations of actions of perception’.61 He agrees with Neurath that ‘a certain connection between the basic sentences and our perceptions is required’, but ‘it is sufficient that the biological designations of perceptive activity occur in the formulation of the methodological requirement concerning the basic sentences … and that they need not occur in the basic sentences themselves’.62 The methodological requirement Carnap refers to here is the stipulation that protocol sentences must be inter-subjectively observable. As far as Carnap is concerned, the requirement of observability is sufficient. So long as the protocol statements are observable, the inclusion of contextualising information is superfluous. And Carnap is not wrong here. For his purposes, such information is superfluous. But as we have seen, for Neurath’s purposes it is far from it.
Carnap’s requirement of observability is a reformulation of his commitment to empiricism (in spirit, his most recent version of verificationism). As such, the requirement of a “certain connection” is in place to exclude the possibility of metaphysical statements entering into the logic of science. But as should now be clear, Neurath’s demand for contextual information is not simply to guarantee such a connection. Neurath’s requirement is not simply a verificationist one. The contextual information itself is of great significance for Neurath, since it is this information that allows decisions on acceptance. Such information however is not necessary for Carnap. As we already saw, his logic of science starts from the assumption that the statements with which he is working are valid. That is what his observability criteria does. Carnap therefore misunderstands the significance of Neurath’s demands for contextual information.
There is one further possible complication here. Neurath’s criteria and Carnap’s requirement of observability are not the same. Carnap argues against Neurath’s protocol statements for having the limitation of being ‘intersubjectively confirmable but only subjectively observable’.63 As Uebel notes, Neurath’s conditions *(ii) and *(iii), about speech-thinking and perception, whilst inter-subjectively confirmable, are not observable by others.64 Nor are descriptions
But before such conclusions are reached, it needs to be emphasised that the requirements on observability placed on observation statements is, as Carnap recognises, a matter of decision about the language best suited to our purposes. The disagreement here is not one over what observability is, but what requirements will be most useful to adopt. Carnap, in line with his principle of tolerance, explicitly frames his choice of observability conditions as a decision.65 What practical reasons are there? That some predicates are only subjectively observable ‘is a serious disadvantage and constitutes reason against their choice’.66 Exactly why this is so disadvantageous is not spelled out in detail by Carnap. Carnap gives no further reasons for rejection. We can accept Carnap’s worries here, but his reasons are far from decisive. The question is then whether Neurath has better reasons than this for embracing inter-subjectively confirmable but only subjective observable predicates.
For Neurath, protocol statements like “Karl’s protocol: Karl formulates: Karl is feeling: Karl is scared” must be potentially valid as protocol statements because of their obvious significance for social sciences like sociology, history and anthropology. ‘Historians of human social life are highly interested in descriptive terms, such as deal with the feeling-tone of persons, their devotion their fear and hopes’.67 Our physicalist language needs to allow for reports of ‘the state of a person who hears Beethoven or looks at certain forms of architecture’.68 If anything, he argues, we need a more extensive and fine-grained terminology for describing feeling-tones. On this point, Neurath seems unarguably right. The disciplines of psychology, anthropology and sociology would
10 Conclusion
A fuller understanding of the eminently practical (if not always clearly expressed) motivations behind Neurath’s protocol sentence schema ought to dissolve any concerns about the practicality of Neurath’s proposal. With regards to Uebel’s bipartite metatheory thesis, recognising the details of Neurath and Carnap’s mature conceptions demonstrates no theoretical disagreements that would undermine its plausibility. In fact, it demonstrates the essential conceptual and theoretical agreement masked by a subtle but significant difference in use of terminology. I also hope to have shown how the distinction between protocol bank and data bank can help us to conceptualise the continuity between the two parts of the meta-theory which can be masked by the division of labour and differences of emphasis that are an inevitable element of a bifurcated project, and the process of protocol acceptance within this.
‘Sentence’ and ‘Statement’ are used interchangeably in this chapter.
Uebel 2007b, p. 435.
The name ‘pragmatics of science’ is taken from Frank (cf. Frank 1957, p. 360), not Neurath, but is less cumbersome than Neurath’s term, the ‘behaviouristics of scholars’ (Neurath 1936, p. 137). ‘Logic of science’ is Carnap’s own term for his project (Carnap 1934).
Neurath 1932, p. 93.
Neurath 1935, quoted in and translated by Uebel 2007b, p. 386.
This historical overview is a slight simplification. Uebel identifies five different positions maintained by Carnap between the Aufbau and Testability and Meaning (cf. Uebel 2007a, p. 442). But the slightly simplified account given here is sufficient for current purposes. For a more detailed history of Carnap’s developing conceptions of protocol statements and methodological solipsism, see Uebel 2007a; 2018 pp. 372–375.
It could be argued that the ideas in Testability and Meaning are a development of Carnap’s talk in 1935, later published as Truth and Confirmation, and that the mature period could therefore be dated from 1935. But only in Testability and Meaning is his conception of protocols stated explicitly.
Carnap 1928, §59.
Carnap 1930, p. 145.
Carnap 1930, p. 144.
Carnap 1930, p. 144.
This is made clear in unpublished manuscripts from 1930. See Uebel 2007a, pp. 191–200.
Carnap 1932a, p. 88.
Carnap 1932a, p. 42.
Carnap 1932b, p. 191.
Carnap 1932a, p. 43.
Carnap 1932c, p. 465.
Carnap 1932c, p. 465.
See Awodey and Carus 2007, pp. 183–192.
See Haller 1982, p. 121.
Carnap 1932c, p. 469.
Carnap 1932c, p. 469.
That this completely conventional conception of protocol statements is not Carnap’s final mature conception is not always made clear in secondary literature. See for example Creath 1990, p. 412; Coffa 1991, p. 371; Richardson 1997, p. 211; 2000, p. S158.
Carnap 1936/37, p. 466.
Carnap 1936/37, p. 455.
For the best examples, see Carnap 1936/37; 1956.
Russell 1941, p. 146.
Carnap 1932c, p. 465.
Carnap 1932c, p. 465.
Neurath 1946, p. 233.
The choice of “synoptic” is deliberate, as it echoes Neurath’s usage of the same terms elsewhere. Neurath describes the purpose of his Encyclopedia as allowing science to be ‘presented to us synoptically in its totality’ (cf. Neurath 1938, p. 141). I think the same motivation underlies the proposed structure for protocol statements.
Nemeth 2019, p. 130.
Neurath 1932, p. 93.
Neurath 1941, p. 226.
Neurath 1941, p. 220.
Neurath 1941, p. 220.
The order of stimulation state and observable fact (Neurath’s D and C respectively) has been switched in my schema. This is based on Neurath’s own usage in a letter to Kaufman from 1935, quoted in Uebel 2007b, p. 386. The advantage of this order over that used by Neurath in 1941 is that it places the object sentence (iv) at the centre of the protocol. The importance of the centrality of the object sentence will be made clear below. For more detailed discussion, see Uebel 2007b.
Uebel 2009, p. 6. In an earlier work, he instead uses the term “content statement”, but I adopt the more recent terminology (cf. Uebel 2007a, p. 388).
Uebel 2007a, pp. 383–388.
Uebel 2007a, pp. 383–384.
This could be understood by analogy to the incommensurability between Kuhnian paradigms. Reports made from within a Newtonian paradigm are corrupted if we attempt to translate or reinterpret them within an Einsteinian paradigm. The markings of the initial paradigm cannot be removed without loss or corruption of the report itself.
This use of terminology follows Uebel. See Uebel 2007a, pp. 384–385.
The above is not intended to capture or provide a specific logic of acceptance. Rather, it is only intended to demonstrate that even the most rudimentary attempt to formalize Neurath’s conditions according to any such logic would obliterate the visually communicative purposes of Neurath’s structure.
Neurath 1932, p. 93.
It seems likely that this confusion is a result of English being Neurath’s second language. In German, “brackets” can be used not simply to refer to the symbols themselves, but the symbols and the clause contained within, as with “parenthesis” in English.
Neurath to Carnap, September 25, 1943 (asp rc 102-55-03), in Tuboly/Cat 2019, p. 596. Importantly, nothing he says here is new.
Carnap to Neurath, February 4, 1944 (asp rc 102-55-04), in Tuboly/Cat 2019, p. 609.
Uebel 2007a, p. 396.
Neurath 1934, p. 107.
The proposed terminology is my own. However, I think the distinction is not only consistent with Neurath’s account, but potentially implicit in his work. For example: ‘A distinction will certainly be made between the protocol statements (that turn up as physical formations) made by an astronomer, or a chronicler, and the statements that have a precisely defined place within a physical system, though obviously there are some overlapping transitions’ (Neurath 1931, pp. 65–66). The former is the protocol bank, the latter the data bank.
Given Neurath’s insistence on the revisability of protocols, it should also be noted that membership in the data bank is also open to revision.
Quoted in and translated by Uebel 2007b, p. 386.
As Uebel notes, this provides Neurath with the additional benefit that ‘it allows for the convergence of reports in different sense modalities to be clearly displayed’ (Uebel 2007b, p. 387). This differentiation allows for the mutually reinforcing use of protocols indexed to different senses supplying the same object sentence.
Carnap 1936/37, p. 13.
Quoted in and translated by Uebel 2007b, p. 386.
Neurath 1946, p. 233.
Carnap 1936/37, p. 13.
Carnap 1936/37, p. 13.
Carnap 1936/37, p. 11.
Uebel 2007a, p. 395.
Carnap 1936/37, p. 9–13.
Carnap 1936/37, p. 12.
Neurath 1944, p. 14.
Neurath 1944, p. 15.
Carnap, 1936/37, p. 12.
References
Awodey, S. and Carus, A. (2007) ‘The Turning Point and the Revolution: Philosophy of Mathematics in Logical Empiricism from Tractatus to Logical Syntax’, in Richardson, A. and Uebel, T. (eds.), The Cambridge Companion to Logical Empiricism, Cambridge: Cambridge University Press, pp. 165–192.
Neurath, O. (1936) ‘Einzelwissenschaften, Einheitswissenschaft, Pseudorationalismus’, in Actes du Congres Internationale de Philosophie Scientifique, Sorbonne, Paris, 1935, Facs. I : Philosophie Scientifique et Empiricisme Logique, Paris: Herman & Cie, pp. 57–64, translated by R.S. Cohen and M. Neurath as ‘Individual Sciences, Unified Science, Pseudorationalism’, in Neurath (1983), pp. 132–138.