Coherentist and Conventionalist Elements in Logical Positivism

The coherentist and conventionalist elements of logical positivism are closely associated with its structuralist elements. The reason for this is that the coherence of a theory depends on its structure rather than its content, and if we are to choose one of many equally coherent theories without any foun-dations this can only happen by means of conventions. These three theories together represent the anti-empiricist tendency in logical positivism.

The tension between the empiricist and non-empiricist elements of log-ical positivism was to some extent addressed already within the movement in a debate between Neurath and Schlick. Unfortunately, this debate was plagued by serious confusions. The epistemological question whether knowl-edge has foundations was regularly confused with semantical questions such the definition of truth. Schlick defended both foundationalism and the corre-spondence theory of truth already in his pre-positivistic period in an article

originally published in 1910 and translated into English in [Sch79b] and later as a full-fledged positivist in [Sch34] (translated in [Sch59]). Neurath on the other hand apparently defended both a coherence theory of justification and a coherence theory of truth (e. g. in [Neu32] - translated in [Neu59] - and in [Neu34]), though there is some dispute about the latter. Of course, these views are not necessarily connected in any obvious way; it can be argued that one can also consistently be a correspondence theorist about truth and a coherentist about justification.

However, this confusion may not matter much, as coherentism is not a plausible theory of either truth or justification; the claim that any arbitrary coherent fairy tale is justified is not much more sensible than the claim that any arbitrary coherent fairy tale is true. While the correspondence theory of truth does not imply that we should be capable of comparing truthbearers with reality, foundationalism does, and Schlick argues in my view quite plausibly that we can in fact compare them with it. Another confusion is more fatal.

Schlick and Neurath both presupposed that foundationalism led to the view that absolutely certain knowledge was possible; they did not see that a weak, fallibilistic foundationalism was possible. See e. g. [Haa09, page 54]

for a distinction between weak and strong foundationalism, where strong foundationalism requires conclusive aka decisive justification, while weak foundationalism requires only defeasible, prima facie justifiation. Bertrand Russell, who had began as a strong foundationalist, retreated in his later philosophy to weak foundationalism (if indeed his theory remained founda-tionalist at all and did not anticipate foundherentism) and defended it in arguing against Neurath, saying in [Rus40, page 150] that

a non-inferential belief need not be either certain or indubitable.

Schlick begins [Sch34, page 79] (translated in [Sch59, page 209]) by ar-guing that all the important attempts at establishing a theory of knowledge grow out of the problem concerning the certainty of human knowledge. How-ever, the real problem Schlick eventually has with Neurath’s view in [Sch34, page 86] (translated in [Sch59, page 215]) is not that it does not guarantee certainty but that it makes the choice of theories entirely arbitrary, so that one would have to consider fairy stories to be as true as a historical report.

A weak foundationalism suffices to make the choice of theories non-arbitrary

and non-conventional although it does not offer incorrigibility. On the other hand Neurath thought falsely that rejecting the arguably implausible posi-tion that human knowledge could be incorrigible required the total rejecposi-tion of foundationalism²⁰. There is then a third option besides the views of Schlick and Neurath.

Extreme conventionalism has also serious problems, just like the coher-entism with which it is often associated. Friedman sees in [Fri91, page 517]

it as a great achievement of Carnap that he extends Poincare’s convention-alism to logic itself. However, that very extension has been revealed to be an especially problematic variation of conventionalism. While without that extension conventionalism is just a corollary of coherentism, this extension makes it a stronger and hence more absurd claim that coherentism. Quine argued in [Qui76] that the view of many logical positivists that even logi-cal truths were conventionally true leads to a vicious infinite regress (even though Quine was in that article willing to entertain the rather extreme view that set theory was conventional). The basic argument against logical con-ventionalism is quite simple. As logical truths are infinite in number, they cannot be stipulated one by one, but must be deduced from some general conventions by means of some logical rules. However, in so doing we are presupposing the very logical rules or truths we are trying to establish by means of conventions²¹.

There is a still greater problem with logical conventionalism. If we as-sume that the transformation rules (i .e inference rules) of a language can be chosen quite arbitrarily, as Carnap does in formulating his Principle of Tolerance according to in [Car37, page xv], we can (as Arthur Prior showed

20These confusions persist in commentators dealing with logical positivism. Coffa still thought in [Cof91, page 254, 363] that fallibilism was equivalent with the rejection of foundationalism and (in [Cof91, page 374]) that accepting Tarski’s theory of truth would in some way count against a coherence theory of justification. Similarly, Misak says in [Mis95, page 93] that the foundationalist is after a foundation of rock, not of shifting sand. This, however, only holds of strong foundationalists, not of weak foundationalists.

Of course, everyone would prefer a foundation of rock. However, it may be that human knowledge just cannot have such a foundation, and a foundation of shifting sand is better than no foundation at all.

21This argument of Quine’s is quite independent of the arguments in [Qui53e] that Friedman saw as ending logical positivism; in [Qui76] Quine sees analytical statements that are not logical as unproblematic, while in [Qui53e] he turns his attack on them.

Indeed, Quine’s views in the two articles may not fit very well together; the infinite regress argument against logical conventionalism is quite similar to an infinite regress argument other philosophers (e. g. Laurence BonJour in [Bon13, page 184]) have used against the extreme empiricism Quine argued for in [Qui53e].

in [Pri60]) introduce a connective tonk, such that by using the transforma-tion rules of that connective we can deduce any sentence from any sentence whatever. We can then admit any sentence whatever, no matter how much it is in contradiction with experience. Here logical positivism collapses into extreme relativism.

I will later return to the discussion of the conventionalist element in logical positivism when discussing Neurath’s version of verificationism; I will show that in Neurath’s treatment verificationism loses all its force and collapses into conventionalism and hence into relativism.