LINGUIST List 13.1354

Wed May 15 2002

Disc: Falsifiability vs. Usefulness

Editor for this issue: Karen Milligan <karenlinguistlist.org>

Directory

1. Julian Bradfield, Re: 13.1351, Disc: Falsifiability vs. Usefulness
2. Dan Everett, RE: 13.1351, Disc: Falsifiability vs. Usefulness
3. H.M. Hubey, Re: 13.1351, Disc: Falsifiability vs. Usefulness

Message 1: Re: 13.1351, Disc: Falsifiability vs. Usefulness

>LINGUIST List: 13.1351 Tue May 14 2002. ISSN: 1068-4875.
>From: "Dan Everett" <dan_everettsil.org>
>Subject: RE: 13.1348, Disc: Falsifiability vs. Usefulness

>Notice that not a single response has yet attempted to deal with the
>serious objections I pointed out from Hull, Hempel, and
>Lakatos. Because to answer those objections would be quite
>difficult. Just so. Falsifiability is not easy to defend. And if it

Being entirely inexpert in philosophy, I have been reluctant to ask a
simple question, but since Dan raises Hempel again, I will ask it:

In what non-trivial sense is Hempel's objection an objection? As I
understand the objection, it is that if S is a falsifiable statement
such as "the moon is made of green cheese" (in my book, Dan's original
example "2+2=4" is analytically true, and therefore not falsifiable),
and N is a blatantly non-falsifiable statement such as "God exists",
then S & N is falsifiable.

This is clearly true, but why is it a problem? The sentence S & N
contains (in an intuitive understanding) both empirical and
non-empirical content, and taking falsifiability as a criterion has
the effect of making "empiricality" propagate through conjunction; if
you took verifiability as a criterion, it would make empiricality
propagate through disjunction. So what? It still doesn't make N
empirical.

There are plenty of other reasons why simple falsifiability is
inadequate as the sole criterion for empiricality, but this particular
objection doesn't seem very salient.

Message 2: RE: 13.1351, Disc: Falsifiability vs. Usefulness

Let me just clarify one final point relating to the distinction I
attempted to draw in my last posting between falsifiability as showing
something to be wrong and falsifiability as a methodological device. I
think that it is quite fine, in fact desirable in some cases, for a
researcher to put his/her cards on the table and say what facts or
arguments would convince them to abandon their hypothesis. Thus if I
say that "in language x stress I believe that stress is always found
on the antepenult" and that I would abandon this view if a word is
found with stress on the penult, then I am accepting/advocating a
methodological point, which, if we so desired, we could label
falsifiability-m. But this is not the same as Falsifiability in the
Popperian sense of showing something to be wrong. I think most of us
do operate with the useful notion of falsifiability-m. And I am quite
happy with this, so long as we recognize that it does not entail
showing something to be wrong or eternal abandonment of the hypothesis
(for example, someone could convince me that there is a word boundary
after the penult in the word that I had thought to be a
counterexample, after I had abandoned my original stress
hypothesis). I urge linguists who wish to hang on to a vestige of
falsifiability to accept this much less potent, non-Popperian sense of
falsifiability, rather than the considerably more questionable and
pompous notion that Hull, Hempel, Lakatos, and other philosophers of
science have argued so strongly against.

Message 3: Re: 13.1351, Disc: Falsifiability vs. Usefulness

A few comments on falsifiability.


> Date: Tue, 14 May 2002 11:25:12 -0300
> From: "Dan Everett" <dan_everettsil.org>
> Subject: RE: 13.1348, Disc: Falsifiability vs. Usefulness
>
>
> falsifiability. To falsify something is to show it to be wrong.

Indeed, if one looks at the assigment of values to the implication
e.g. P=>Q, the only time it is assigned "false" is when P is true and
Q is false. This is obviously the counterexample. If the claims is that
"all rabbits are yellow" then finding a rabbit that is not yellow (the
counterexample) falsifies the statement.

There is no principle upon which it can be based. Anyone and everyone
can see that this is true merely from living in the world.


> Sharifian also says that: "Even if "one
> simply finds no (or few, if Gale is correct) examples of
> falsifiability playing a role in acknowledged advances in science",
> which is definitely an oversimplification, still it doesn't harm the
> LOGIC." This fails to respond to the last point of my last letter,
> namely, that either sciences advances illogically or falsifiability is
> not logically necessary to the advance of science.

Logic is not the only mode of reasoning. There is also Bayesian
reasoning. For example, in 1865 James Clerk Maxwell, while
playing around with the known equations of electricity decided to
add a new term to one of the equations because it made things
symmetric and it looked better that way. Then he manipulated
them mathematically and derived the wave equation. When he
computed the speed of the wave and saw that it was equal to the
speed of light, he decided that light was an electromagnetic wave and
that there existed electromagnetic waves. In the 1890s Heinrich
Hertz made an apparatus that would generate such waves, and
an apparatus that would receive such waves, sent such a wave
and received it. About 30 years later we got the radio. Obviously
it might not have worked. If he did everything correctly and he
could not receive the waves, Maxwell would have been falsified.

But look at Maxwell's idea differently. Speed of light is 3*10^8
meters per second. What is the probability that EM waves would have
had that speed? After all, at that time we had no idea of such
limits. Maybe EM wave speed was 10^100 or 10^200. Suppose
Maxwell was off by a factor of 100, e.g. 3*10^6 or 3*10^10. The
difference between the two is on the order of 10^4. If we divide
this by 10^100 we get 10^(-96). That is 0.000....1 (with 96 zeros
in the front of the 1). The number of elementary particles in the
universe is 10^87.

So the probability of getting the speed of EM waves to be around
the same speed as that of light (even off by 100 times) is
almost zero. And being off by 100 times is really bad in science.
For example, if my salary were 100 times what it is now, I might
not have been reading this email at all :-)

So even if Hertz' apparatus did not confirm Maxwell, physicists
would have tried something else again because the odds are too
small that the speed of EM waves would come to be in the
neighborhood of the speed of light.

But this reasoning is essentially probabilistic (more Bayesian), but
the simpler version (e.g. the logical one, falsification) still applied
in the case of Hertz' experiment. If everything had been done
according to theory and nothing observed there would have been
a falsification.


> Therefore, and this
> is vital from a pragmatist point of view, the discussion has failed to
> unearth an clear exemplar that falsifiability makes a difference to
> practice.

Maybe those who keep practicing are unaware of the concept
of falsifiability.


> Finally, Denis Bouchard's remarks. Consider his rebuttal on my urging
> that we consider usefulness as an alternative to falsifiability:

There is no reason why "useful" models (especially if mathematical)
should not be used. The main problem in the social sciences is that
the theories are too grand for the limited evidence. It is too complex
for a simple theory to be true especially since many of them are
based on very fuzzy concepts. The conclusions are too far-reaching,
and the evidence slim.

> Ultimately what I wanted to do in beginning this discussion is just
> this - to show that falsifiability is a dubious notion, not nearly as
> straightforward as linguists all too often think.

What is really dubious are theories of linguistics, and social sciences.
Falsifiability is simple and straightforward. Usually the chain of
reasoning in social sciences is too simple for the complex problem
at hand.


> From: Robert Whiting <whitingcc.helsinki.fi>
> Subject: Re: 13.1334, Disc: Falsifiability vs. Usefulness
>
>
> Falsifiability lies in the simple fact that a proposition and its
> opposite (or P and ~P) cannot both be true at the same time (although
> both may be false).

That is certainly not possible. P + ~P =1 always. If P=0,
and ~P=0, then we'd have P+~P=0 which is not possible.



> From: Mark Douglas Arnold <mdarnoldwam.umd.edu>
> Subject: Re: 13.1348, Disc: Falsifiability vs. Usefulness
>
> Though the following might seem flippant, it is not meant to be.
>
> Dan Everett, in 13.1334:
>
> "So either science progresses illogically, or falsifiability is not part
> of its logic."
>
> Presumably Everett takes the apparent absurdity of the first proposition
> to set up the coup de grace in the second. I humbly submit that the first
> proposition is not as obviously false as many of us (seem to) want to
> believe.

I assume that logic (in "illogically") refers to only one way of reasoning,
that of formal logic. There are other means of reasoning, e.g. Bayesian,
probabilistic etc. But they differ in some details. Bayesian "logic" like
fuzzy logic could handle statements such as "probably true", or
"mostly true" or "most likely true" whereas logic cannot.

Logic is kind of biased in favor of truth. For example, for the implication
P=>Q, for the case P=0, and Q=0, the assignment to the implication is
"true". But why? If P=>Q is "All gazooks are kazooks", and if we have
seen no gazooks and no kazooks why would we assign the value "true" to
this specific case?

A little history here. Hempel's "Raven Paradox" was the nail in the coffin
of the "confirmation/verification" theory of science, e.g. one gets a
hypothesis, and then
conducts and experiment and the experiment confirms or verifies the hypothesis.

Hempel's argument was that, if P=>Q is "All ravens are black", then every time
we see a raven that is black, we "confirm/verify" the statement. But P=>Q is
also equal to ~Q=>~P (If it is not black, it is not a raven), so then
everytime
we see a thing that is not black (e.g. yellow banana, red corvettee) we are
"confirming" that all ravens are black.

So the only thing left was "falsification".

In order to falsify the statement "all ravens are black" you simply find
a counterexample, e.g. find a nonblack raven.


-
M. Hubey

hubeyhmail.montclair.edu