[hist-analytic] Reichenbach, Carnap, Positivism

Steven Gimbel sgimbel at gettysburg.edu
Mon Sep 28 14:40:01 EDT 2009

“What motivated Reichenbach, and the Berlin positivists, was the idea that truth is not a very good concept for science to work with. Nothing is known with certainty to be true. What is known is what is probably true.”


I think this is partially right, but there’s more to the story – more in two parts:


Part I:


While Reichenbach and Carnap are both interested in rational reconstruction of scientific theories, what this reconstruction consists of differs importantly between them.  Reichenbach is much more interested in theory change, in theories as part of an historical process, and wanting to show the development in terms of axiomatizations. 


I think the best way to interpret his first two books is attempts to model epistemological change on change in physical theory.  The move from Newtonian to Einsteinian mechanics not only presents us with an opportunity to rethink epistemic foundations, but also provides us with the template through which to understand not only this theory change, but the process in general.  Just as Newton’s results are first order approximations of Einstein’s, so too the concepts ought to be seen as “first order epistemic approximations.”  


In the place of Taylor series expansions, we put a sort of neo-Kantian analysis that differs from Kant in that the synthetic a priori propositions are constitutive, but not apodictic.  We need theory to make sense of the raw manifold of perception, but these theoretical postulates are revisable in light of observations. This is scientific progress – replacing one sets of axiomatic assertions with another whose empirical basis is more inclusive. 


When we do revise them, when we have a new theory take over for an old one, we start from the old concepts as our starting point, we cannot rebuild from pure observations in a protocol language.  As such, the new theory is still pregnant with the old to some degree.  In this way, the process is cumulative and progressive, but not completely revolutionary.  We know that the new theory’s axiom set with its basic concepts will also be replaced eventually. This, I believe, was one of Reichenbach’s motivations – he saw the error of Kant not only as tying his epistemology to Newtonian mechanics which was overthrown, but more generally tying it to any given theory.  He wanted to draw out the lessons of the rise of relativity theory without binding his view to the dictates of the theory in the same way.  Thus, talk of truth was a throwback to the Kantian apoctic claims – exactly what did in Kantianism in his view – and must be avoided.  Rather, scientific theories in the move from Newton to Einstein are getting better, and hence we need to speak in terms of degrees, shades.  This requires abandoning talk of truth for talk of probabilities.


Part II:


Here, I’m pulling from the wonderful work of Flavia Padovani, whose scholarship on probability and causality in Reichenbach’s early writings is well worth the time.  


Reichenbach’s pre-relativity work (especially his dissertation of 1915) focuses on the relationship between probability and physics.  He was an engineer before turning philosopher and so he appreciated the pragmatic as much as the theoretical.  As such, the advances in statistical mechanics fascinated him.  Here the theoretical instrument of statistics turned into the generator of actual physical laws.  That’s weird and seemed to indicate something deeper.  


Independent of statistical mechanics, though, the same lack of absolute determination was at play in anything that required measurement, and all scientific theories begin with measurement.  There is therefore a probabilistic element underlying all scientific investigation.  How to relate the principles of probability and the principle of causality within an intellectual framework that supports laws of nature became the central question of epistemology for him.  As such, the move away from truth to something necessary.


If you combine his early interest in the role of probability with his interest in concept change from Einstein’s seminar at Berlin, the rejection of truth talk for probability talk seems quite natural.  Well, to some degree…






Steve Gimbel

Chair, Department of Philosophy

Gettysburg College


blog: Philosophers' Playground <http://philosophersplayground.blogspot.com/> 


From: hist-analytic-manager at simplelists.com [mailto:hist-analytic-manager at simplelists.com] On Behalf Of Baynesr at comcast.net
Sent: Saturday, September 26, 2009 11:56 AM
To: Roger Bishop Jones
Cc: hist-analytic
Subject: Re: Reichenbach, Carnap, Positivism


"As soon as you attempt to describe the sense data you will
introduce the possibility of error,"


So I can be in error in describing them, say thinking the hen sense datum has five instead of its in fact four speckles? How about being in error in believing there is a datum? If you can have one sort of error, why not the existential?


"(this is like the skeptics "appearances appear")"


Some would hold, following Plato in the Sophist, that there are no appearances because there is no reality; there is just the way we describe our experiences; "reality" need not apply.


There are a number of places where you, rightly, point out I haven't said enough; but these posts as you can see are very long. I think you should take a look at some of this stuff, time permitting.

"I am generally not in favour of resort to talk about probability
as a remedy for uncertainty, or the various other similar
stratagems (such as confirmation theory)."


Some deny there IS truth, only that there are degrees of likelihood of being true; or something close to this.

A cynical comment on my part on Hume: If Hume is right, there are a few conventions having to do with arithemetic or vacuous tautologies at best, and then there is the psychology of belief. There is no real need for philosophy, unless we think of it as a way of describing psychological facts of experience. An overstatement? Just a little, perhaps.

> This will effect
> the logic of science. Truth values will be infinite between 0 and 1. There
> won't be two truth values. Much of this was supported by the Heisenberg
> business.

I think this is a bad idea.

Cmon Roger. First you say I don't say enough and then on an idea that prevails from Reichebach to Zedeh you say its a bad idea. I think it is a pretty darn good one, especially if it allows us to exit the domination of trivial problems raised by the Tarskians.


What I mean by a "theory of ontology" is a theory as to how to answer the question "What is there?" in the sense that epistemology answers the question: "How do I know?" (in a broad sense of 'how', including what is knowledge etc

In the Aufbau, Carnap is pretty much a phenomenalist. I think you are relying too much on Schilpp, but I may be wrong.


By the two language approach, I mean Carnap's reliance on the protocol language on the one hand and the language of science on the other.


I can't answer your "I don't understand" questions. I need to know what exactly I say that you don't understand. You quote a long paragraph. I am reluctant to write another one of greater length for fear of not being undestood. I need more specificity.


If mathematics is a bunch of tautologies and explanations are of the sort Carnap accepts (the sort of thing Hempel had in mind, i.e. deductive nomological explanations etc) then the question is what is the basis for thinking these empty sentences relate to the world? Reichenbach's answer is that this is owing to their use in making predictions. Otherwise they say nothing of value. I'm not sure what you mean by a possible world. I never been to one.

"Someone is surely a materialist only if he asserts that only matter exists,"


Yes, and a physicalist who believes what is physical is material is a physicalist just the same, and conversely a materialist who believes that matter is physical is a materialist just the same. No problem here. Language has little to do with it.


A language containing theoretical terms is "reducible" through operational definition to the physicalist language, so physical language and theoretical language differ only in the descriptive constants being in the one case terms that refer to unobservables. So there is no difference in languages, especially once the protocol language, via Neurath, is reduced to a physical language.


I missed a couple of questions. They would require new books to anwer thoroughly and I'm still trying to read the old ones.






----- Original Message -----
From: "Roger Bishop Jones" <rbj at rbjones.com>
To: Baynesr at comcast.net, hist-analytic at simplelists.com
Sent: Friday, September 25, 2009 4:02:44 PM GMT -05:00 US/Canada Eastern
Subject: Re: Reichenbach, Carnap, Positivism

My last message was written in too great a haste so there are some points
of retraction here.

On Monday 21 September 2009 14:32:15 you wrote:
>The connection between Descartes and the early protocol language advocates
> was certainty. In each case, you were dealing with a contingent fact that
> was known with certainty: that I exist is a contingent fact and, yet, I
> know this with greater certainty than I know the truth of the
> multiplication tables; similarly, 'Red, here, now' is a contingent fact I
> know with certainty. That was the reasoning, anyway. The comparison is
> quite strong, actually. I've never been completely convinced that the
> Cogito is a bad argument.

Actually, I don't have a problem with the cogito.
It is as you suggest, a sound argument from a contingent premise,
(except perhaps for Leibniz, who would regard the premise as
necessary but the conclusion contingent).
We might possibly disagree about whether it is a priori.
The premise is pretty close to indubitable for the "I" involved,
but not for anyone else.

> In fact, it is a better argument in some ways
> than the idea that we have certainty with respect to knowledgeo of our
> sense data. So I'm more inclined to go with Descartes. Is there something
> specific about Descartes argument that you think is wrong.

The cogito is OK.  Please disregard my objection.
Also, I am not now inclined to say that we have knowledge of sense data,
I think we just have the sense data.

In the kind of theory of perception which I might be inclined to
support, sense data are just that, data, rather than propositions.
They are evidence on which our "inferences" about the external world
are based, and they are incorrigible simply because they are what
they are, and are not actually propositions.
As soon as you attempt to describe the sense data you will
introduce the possibility of error, but this is not a part of 
the process of perception, so this kind of error is arguably
epistemically irrelevant.
I think this means that I am disinclined to say that sense data
are corrigible, but also that it is not quite right to say that
they are incorrigible either.  They are what they are.
(this is like the skeptics "appearances appear")

>I'm unclear on what your point is with respect to probability.

Looking back, I think I spoke too hastily on this (as well).
You didn't say enough about why Reichenbach decided that we
should talk about probability rather than truth.
I am not sympathetic to that proposal, but don't know enough
to go further than that.

I am generally not in favour of resort to talk about probability
as a remedy for uncertainty, or the various other similar
stratagems (such as confirmation theory).

> What
> motivated Reichenbach, and the Berlin positivists, was the idea that truth
> is not a very good concept for science to work with. Nothing is known withe
> certainty to be true. What is known is what is probably true.

It certainly was my intention to challenge this proposition.
If the doubt about certain knowledge of truth is Humean
(in which case I think it would be more accurate to say that
no contingent proposition can be known demonstratively to be true),
then Hume argues correctly that the same can be said about
probability claims.  In your words, nothing is known with
certainty to be probably true.  In my explication of Hume,
claims about the probability of contingent propositions
are not demonstrative either.

> The merit is
> in the simple fact that science is not a body of propositions known to be
> true; it is a body of propositions thought to be probable.

If you weaken it to "thought" then you can probably forget the
"probable" for I think most people do think that the propositions
of science are true.

> This will effect
> the logic of science. Truth values will be infinite between 0 and 1. There
> won't be two truth values. Much of this was supported by the Heisenberg
> business.

I think this is a bad idea.

>By the way, the sense data program as an theory of ontology is very
> different from the sense data "program" of the positivists.

I don't understand what you mean by "theory of ontology" here.
(Or how the sense data program can be one).

> Some positivists were realists, others were phenomenalists.

And Carnap was neither.

> In the early going,
> when Mach was a major player, phenomenalism was preferred because it linked
> up with verificationism and operationalism in physics. There were actually
> two historical strands: one dealing with the aftermath of the Special
> Theory and the other having to do with criteria of meaningfulness and the
> status of metaphysics - pseudo-problems and all that.

>"I think this is a mistake, and I would guess that this is a point
>which Carnap never conceded."

I think you should have included a bit more context there.
That sentence (of mine) was a response to:

>>Interestingly, Reichenbach notes that it was the issue of probability that
>> distanced the Berlin logical positivists from the Viennese logical
>> positivists. The Berlin people thought that *prediction* could not be
>> addressed within the framework of a logic that reduced to tautologies.

>First off, I think he should have conceded. In a way he did. The protocol
> language became a physicalist language. But this threatened the relevance
> of the two language approach. All we really need is the physical language.
> The rest is "inventory."

I don't really understand what you refer to here as
"the two language approach".
Is this a reference to his persistent interest in relating the 
languages of science (in some way) to observation statements?

Carnap was a pragmatic pluralist, and this doctrine is articulated
in his "principle of tolerance".
This meant that he was prepared to accept physics using the language
of physicists (or a formalisation of it), and sought to understand
the relationship between such languages and observational descriptions.

>I don't think the historians have figured out what was in the back of
> Reichenbach ingenious mind. As I see it, here is the situation. Reichenbach
> had seen that in dealing with the alternative views of space available,
> Riemannian, etc, that what you had was basically a vacuous formalism UNTIL
> you selected a definition of 'congruence'. Without this no geometry was
> testable, but more importantly the concept of 'distance' without congruence
> was meaningless. So you had these "analytical" truths of geometry that had
> no physical significance whatsover as long as this notion was not clearly
> defined. Now in the case of analytic sentences understood in the narrow
> sense as a bunch of tautologies you had much the same situation: sentences
> that tell us nothing about the world. So how do you get these sentences to
> be relevant without falling into the synthetic a priori? Answer: you say
> that what is essential is linking the tautologies to the world by way of
> the semantics of sentences at the basis of science: predictions being the
> most obvious. So, if I'm right, tautologies stood to predictions as
> geometry stood to congruence, to Reichenbach's way of thinking. Again, he
> was driven to much of this by way of a rejection of synthetic a priori.
> This is why I said this idea and not analyticity was what was important.
> Not much hangs on analyticity, in my opinion. Much hangs on the synthetic a
> priori, especially in physics.

That sounds rather different to Carnap's position, as I understand it.

I'm afraid it doesn't make a lot of sense to me.
I don't see why there is a problem in analytic sentences not telling
us anything about the real world (at least in the sense of not
excluding any possible world), and I'm not at all clear why
you want to link tautologies to the real world or what you mean by that.

On the matter of the synthetic a priori being more important than the
notion of analyticity I am also a bit baffled.
The whole discussion presupposes the concept of analyticity.

>On contexts of justification and discovery, my view is this; to dismiss the
> circumstances of discovery is a "cop out."

I don't think I would advocate dismissing the circumstance of discovery.

> The history of science is an
> appropriate object of scientific investigation. So on my view saying: "Oh,
> I just dreamed this whole thing up in a nightmare" (something like Kekule
> with the carbon ring) that is just to defer investigation of possibly the
> most interesting types of events known to science: scientific theorizing
> itself. This is just more "sweep it under the rug" methodology; the sort
> that was encouraged by the verificationist theory of meaning.

But surely, this kind of thing really does happen?
Though not out of context.
A mathematician may come up with a solution to his problem in a dream.
What is swept under the rug in admitting this?

Tell me also how this connects with the verificationist theory of
meaning.  (I don't endorse it, but it does seem to me a constructive
idea rather than a "sweeping under the rug").

>On Carnap's materialism: As long as the language of science is
> physicalistic, which for Carnap it was, and as long as you believe that
> science investigates reality, it is physicalistic, to that extent Carnap
> was a physicalist.

That sounds rather more definite to me than Carnap.
Carnap thought that the "theoretical" language was the one closest
to the language of science (physics), not the physicalistic language.
In any case he admitted all these languages.

Someone is surely a materialist only if he asserts that only matter exists,
and Carnap did not do that.
To call him a physicalist I think you would have to establish that he
has some definite preference for physicalistic language, but I have
seen no sign of this.

> Indeed, if you probe more deeply his 'intensions' may
> prove to be physical, although I think that you are right to point out the
> possibility of an 'anamoly' in his thinking. By the way, his "theoretical
> language" IS a physicalistic language!

I don't know what you mean by that, it seems clear that Carnap considered
physicalistic and theoretical language distinct, though quite possibly
the latter is an extension of the former.
I know of these different languages only what he says in his intellectual
autobiography, which is very little.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rbjones.com/pipermail/hist-analytic_rbjones.com/attachments/20090928/a3fbcd10/attachment-0002.html>

More information about the hist-analytic mailing list