LINGUIST List 13.1402

Sun May 19 2002

Disc: Addendum/Falsifiability vs. Usefulness

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


Directory

  1. Dr Stefan Ploch, RE: 13.1367, Disc: Falsifiability vs. Usefulness
  2. Dan Everett, Reply to Ploch

Message 1: RE: 13.1367, Disc: Falsifiability vs. Usefulness

Date: Sat, 18 May 2002 23:52:34 +0200
From: Dr Stefan Ploch <plochlanguages.wits.ac.za>
Subject: RE: 13.1367, Disc: Falsifiability vs. Usefulness

Editor's Note: Because Stefan Ploch's submission was "trapped" in the
Linguist system and was not discovered until after the discussion was
closed, we are posting his message. Since Dan Everett initiated the
discussion, he was given the opportunity to reply. His response is the
second message in this issue. There will be no futher postings on this
topic.

Stefan Ploch writes:

Having followed the discussion on the Linguist List re. falsifiability
vs. usefulness, it appears to me that Dan Everett argues against a
version of falsifiability that Popper did not only not have in mind
but which Popper also explicitly excluded from his investigation (cf.
below for details). Apart from that, falsifiability (in Popper's
sense) is the only arguable objective criterion, as far as I know the
literature. Falsifiability is actually, and as some have pointed out,
about a logical relationship between an assumption and its
predictions, it is an alternative to the logical fallacy to think that
assumptions can be justified by providing examples where they work. So
falsifiability has to do with objective truth and objective knowledge.
Whether Popper is correct in his logic, and it appears that he is, is
one matter; whether anyone finds that useful, is another.

If I understand Dan Everett's views about the supposed non-usefulness
of Popperian falsifiability correctly, then the following two
statements by Everett should provide a good summary of them
(abstracting away from certain details for the moment):

(1) "But this just means that falsifiability is nonlethal to
hypotheses, i.e. that it can always be circumvented, which makes it
less than useful."

(2) "To falsify something is to show it to be wrong. It is
not merely to convince someone that another way of doing things
handles the facts better than their original approach (that is
usefulness)."

Re. claim (1): Popper does not claim that any hypothesis can be
demonstrably falsified; he in fact says that it cannot, because one
can always save any hypothesis by an ad hoc hypothesis (in Everett's
terms, "it can always be circumvented"). Popper made it very clear
that there are (at least) two meanings one could assign to
falsifiability: falsifiability1, which refers to a logical
relationship between a universal statement and the class of its
potential falsifiers, and falsifiability2, which refers to the
question whether a hypothesis can be conclusively falsified (i.e,
whether it is demonstrably falsifiable). Popper also stressed that he
is referring to falsifiability1. It is in this context that Popper
points out that any hypothesis can be circumvented (by a so-called
'conventionalist stratagem', which is, more or less, an ad hoc
hypothesis), because of which no hypothesis is in his view
'demonstrably falsifiable' (=falsifiable2). Consequently, there is no
argument in attacking Popper by saying that falsifiability, in the
sense of demonstrable/conclusive falsifiability (= falsifiability2),
is not useful (because it can always be circumvented): Popper made it
quite clear, as early as 1934, that the meaning of falsifiability he
makes use of (=falsifiability1) is not the one Everett is referring to
(=falsifiability2); Popper stressed that he is referring to a logical
relationship.

To support my analysis of Popper's views, let me provide the relevant
references: As indicated, Popper distinguished 'falsifiable'1 from
'falsifiable'2; in order to avoid any confusion, he only used
'falsifiable' to denote 'falsifiable'1, and used 'falsified' for
'falsifiable'2. Falsifiable (=falsifiable1) refers to a "purely
logical concept" (Popper 1983: xxii), as described above, falsified
(=falsifiable2) refers to the state of affairs where a theory can
"*definitely* or *conclusively* or *demonstrably* be falsified
('demonstrably falsifiable')". After providing these explanation (and
having pointed out that the same distinction was made as early as in
his Logik der Forschung from 1934), Popper (1982: xxii-xxiii) notes:

"An entire literature rests on the failure to oberve this distinction.
... (Instead of distinguishing the two meanings -- 'falsifiability1',
the possibility // that certain theories can in principle be
falsified, because they have some potential falsifiers, and
'falsifiability2', the *always* problematic possibility that a theory
can be shown to be false [I think Popper means 'cannot be shown to be
false'; the context and further quotes, cf. below, make this clear,
SP], since final empirical proofs do not exist -- the ironic
distinctions of 'Popper0', 'Popper1', and 'Popper2', and so on, have
been made (i.e., of various stages of 'Popper' that flagrantly
contradict one another and cannot be brought into harmony). [There is
a footnote in which Popper refers to a paper by Lakatos' which I will
discuss below, SP.] And the difficulties, in many cases the
impossibility, of a conclusive practical falsification are put forward
as a difficulty or even impossibility of the proposed criterion of
demarcation."

The relevant point here is that it is not a problem for
falsificationism that sometimes (or even often) researchers could in
principle use an ad hoc hypothesis to save an otherwise falsified
hypothesis (because of which no hypothesis is falsifiable2 =
demonstrably falsifiable). It simply follows from Popper's views on
falsifiability that one must as a scientist avoid conventionalist
stratagems (which save an otherwise falsified hypothesis via an ad-hoc
hypothesis from being falsified). In this way, falsifiability is an
idea-checker, and as that it is methodology highly useful. Popper (p.
xxiii) then continues (I am leaving out one paragraph):

"It should be stressed that the uncertainty of every empirical
falsification (which I have myself repeatedly pointed out) should not
be taken too seriously (as I have also pointed out). Then, starting on
the next page (pp. xxiv-xxv):

"The misunderstood logical-technical meaning of falsifiability in the
first sense, in the sense of the criterion of demarcation, has led to
two historical legends. The first, unimportant legend is that I
overlooked the non-conclusiveness of the falsifiability of theories --
the fact that theories are never conclusively falsifiable in the
second sense. Whereas in fact, I had repeatedly stressed this since
1932. The second legend (and it is a far more important legend) is
that falsification plays no role in the history of science. In fact,
it plays a leading role, despite its non-definite character."

Popper then provides the "List of Examples Chosen Almost at Random"
(pp. xxvi-xxx). It seems to me that the Popper Everett is talking
about never existed.


Re. claim (2): Everett says "To falsify something is to show it to be
wrong." No, and as should have become clear from the quotes above, not
when we are talking about what Popper is talking about. Everett is
talking about demonstrating that a hypothesis is wrong, i.e., is
demonstrably falsifiably (=falsifiable2). Popper talks about a logical
relationship (=falsifiable1). Furthermore, when Everett says that "It
[i.e., to falsify something] is not merely to convince someone that
another way of doing things handles the facts better than their
original approach (that is usefulness)", he is actually quite close to
what Popper's method looks like (if we neglect that Popper's views are
non-subjectivist and are not interested in what will convince people):
the concept that one hypothesis is better than the other is precisely
what Popper referred to as 'a higher degree of verisimilitude', which
follows from his trial-and-error method; that which comes closest to
the truth (by having shown that it is least wrong, providing we avoid
ad hoc hypotheses) is, at present, the best hypothesis (which may
change again).

Independently of the discussion on the Linguist List, the distinction
between these two versions of falsifiability is an important one
because it makes clear that falsifiability (=falsifiability1 =
Popper's falsifiability = what I will mean henceforth) does not tell
you anything about which of your available options to take when a
theory is either untestable or falsified but it does give you a useful
methodology with useful options. The available options I have in mind
are these: If a hypothesis is untestable, you can either investigate
another hypothesis or theory (i.e., 'do something else'), or try to
find a test and in this way turn your untestable theory into a
testable one; if it is testable but falsified, you can again change
your solution quite radically, or find an auxiliary hypothesis (with
independent evidence, which you again need testability for!) and in
this way increase the truth content of your theory. Falsifiability
(remember: falsifiability1, the logical relationship) is a checking
device, and certain methodological considerations follow from it (in
short, the trial-and-error method, but Popper is more sophisticated
than that, cf. Popper 1972: 215ff). So a falsifiability-based method
gives us certain options, as specified, but it is not an advisor that
tells people what to do or which of their available options to make
use of. In the end, it is a matter of what one is interested in,
whether one wants to 'drop' a falsified hypothesis or a hitherto
untestable one, or rather continue to work on it; as long as one keeps
checking the falsifiability status of one's theories, there is no
problem. The problem starts where people pretend that they have
evidence by pointing to cases where their hypothesis works and
ignoring cases where it does not or by using ad hoc hypotheses to save
other testable hypotheses from being falsified. Falsifiability will
not tell any researcher what they should be interested in or which of
the options mentioned to go for.

Furthermore, as has already been pointed out in the present discussion
on the Linguist List, usefulness is a subjective criterion, because of
which I reject it completely, as far as the question is concerned what
a given hypothesis is worth, objectively/scientifically speaking. Of
course, what is useful to me may have a great influence on my decision
which of the available options mentioned above (and checked and sorted
by falsifiability) I will actually take: do I try to find a test, do I
'ditch' the hypothesis, do I try to come up with an auxiliary
(independently motivatable, non-ad-hoc) hypothesis, for example? There
is a methodologically important difference between the objective
criterion falsifiability and the subjective (or even sociological)
criterion usefulness: While somebody who insists on testability,
cannot, on logical grounds, maintain that their analysis is closer to
the truth than another one and in this way worth its salt, a supporter
of usefulness can always say that they find something useful. A
subjectivist can always be 'right', and what they say can always be
'useful'. This makes discussion difficult.

A few more points of clarification about falsifiability and what
Popper and Lakatos (whom Everett mistakenly refers to in support of
his anti-falsificationist point) actually said are also in order. I
will start with some general items, and then proceed to provide a
number of quotes by Popper and Lakatos (interspersed with some
comments on my part). Popper pointed out that falsifiability is not
about what people DO (or what Kuhn may have referred to as 'normal
science') but about objective knowledge (as something belonging to
'world 3', the world of objective thought contents). Let us not forget
that people who are employed as scientists or academics may well not
be particularly scientific; similarly, what is done by those who
consider what they do scientific or useful may not be particularly
scientific either. As was mentioned in this discussion by Jens
Stengaard Larsen: "Well, if the task is seeking the truth, then
falsifiability _is_ rather useful". Here is a challenge for all
supporters of usefulness:

Come up with an alternative OBJECTIVE criterion for science (by
science, I am not referring to 'being a scientist' but -- for what is
a sound argument). (For a more precise definition of what I mean by
'objective', cf. Popper's discussion of subjectivism in chapter 2 in
Popper [1972] 1973.)

It can not be stressed enough (a) that whether Popper thought that the
theories of most scientists are actually scientific in his sense is
not in any a priori relationship with whether Popper's falsifiability
criterion works, and (b) that in order for Popper's falsifiability
criterion to 'work' it does not have to useful to anyone; Popper's
criterion works if it can be decided on the basis of logic which of
two assumptions is closer to the truth (Popper's falsificationism is
actually more sophisticated than that, new theories should, for
example, also lead to new problems, new ways of looking at things,
etc., cf. Popper 1972: chapter 10 for details).

Since in the current discussion of falsifiability and usefulness on
the Linguist List it has not been mentioned just how detailed and
sophisticated Popper's methodology actually is, while Popper is
usually just dismissed as 'naive' falsificationist (not by everyone),
and since Popper's views on usefulness and related notions (e.g.,
instrumentalism) have been ignored let alone countered, I would like
to take issue with this misrepresentation and negligence and provide
some of the missing details. I will start by briefly mentioning a few
'Popperian pointers' about three types of usefulness, subjective,
social, and applicational usefulness:

Applicational usefulness is, according to Popper, a concept akin to
instrumentalism. It would therefore have been helpful if Everett had
discussed Popper's views on instrumentalism. The reason for why
instrumentalism is interesting here at all is that, since Everett
attacks Popper's falsifiability, he should argue against the views
presented and argued against by Popper that come closest to the views
that he (Everett) is in favour of, i.e., usefulness � la William
James. A quick check in the "Index of Names" in Popper [1963] (1972)
led me to Popper's presentation and deconstruction of instrumentalism,
cf. chapter 3, "Three views concerning human knowledge", of which
James' 'usefulness' is according to Popper a particular version, cf.
p. 109n24. Similarly, in Schilpp's (1974) index there are a number of
references to William James. Since this monograph contains various
criticisms of Popper's views in combination with replies to them by
Popper, this countering of Popper's counterarguments would have been
most relevant to Everett's point.

Individual/subjective usefulness: In short, the main thing to be said
against subjectivism, and it was said by Popper, is that it is not
arguable; an objectivist rational view is. For more elaborate
arguments against subjectivism, take mostly any book by Popper, and
look in the index under 'subjectivism'; chapter 2 in Popper [1972]
1973, "Two faces of common sense: an argument for commonsense realism
and against the commonsense theory of knowledge", may be useful in
this context.

Social usefulness: What is useful to an individual is logically just
as unrelated to rationality and truth as what is useful to whole
groups. In this sense, an individual version of usefulness (some form
of subjectivism, cf. above) is, from a Popperian point of view, just
as pseudoscientific as a sociological version; the distinction between
individual versus sociological usefulness is irrelevant to Popper as
far as the relevance of his testability criterion is concerned (and
not by Popper's subjective choosing, but because there is no
objectifiable logical relationship). However, for those who want to
know more about Popper's views on social usefulness, consider the
following quote, which provides a kind of spot light:

"Kuhn and I agree that astrology is not a science, and Kuhn explains
why from his point of view it is not a science. [Remember that Kuhn
talks about the psychology of research, SP.] This explanation seems to
me entirely unconvincing: *from his point of view* astrology should be
accepted as a science. For it has all the properties which Kuhn uses
to characterize science: there is a community of practitioners who
share a routine and who are engaged in puzzle solving. ('Puzzles' are,
so far as I can understand, minor problems which do not affect the
routine, or the beliefs linked up with the routine.) Once this has
been seen we may find, in a couple of year's time, the great
foundations supporting astrological research. From Kuhn's sociological
point of view, astrology would then be socially recognised as a
science. This would in my opinion be only a minor disaster; the major
disaster would be the replacement of a rational criterion of science
by a sociological one." (Popper, in Schilpp 1974: 1146f)

Let me now provide some details regarding what Popper's views on some
of the issues relevant here actually are, views which are
unfortunately not discussed by Everett. I have already pointed out
that Everett does not investigate Popper's argumentation against
instrumentalism and usefulness. Similarly, Everett does not mention
that Popper was at pains to distinguish quid juris?, i.e., questions
of logical, epistemological and methodological validity, from quid
facti?, i.e., questions of factuality: "... it was Hume's great
achievement to break this uncritical identification of the question of
fact - quid facti? - and the question of justification of validity -
quid juris?" (Popper [1963] 1972: 45). So Everett does not only
conflate a logical relationship with decision-making tools when he
claims that falsifiability is not useful for 'picking a theory' or for
explaining which theories get picked by researchers -- by which he
also plays down the usefulness of tests and the trial-and-error
method --, he also ignores what Popper, whom he dismisses, thought
about what falsifiability is about and what not. In my opinion,
Everett's argumentation against Popper would have gained substantially
if he had taken into account Popper's distinction of quid juris and
quid facti and then tried to counter falsifiability.

The conflation of quid juris and quid facti AGAINST Popper's explicit
distinction, is older than Everett, and the most famous instigator of
this approach is Feyerabend, who spent whole volumes (cf. 1975, 1987)
on dismissing testability and method by pointing out that testability
is supposedly not what people who are called scientists DO. It is
obviously not a sound argument against falsifiability to attack it on
the grounds of it not being able to have properties it was not set up
to have (e.g., subjective usefulness), or in relation to notions about
which it is not clear why they would be in any a priori relationship
to falsifiability in the first place (again, usefulness, or
applicability in commercial products, or what Galileo Galilei did and
why, etc.). What linguists do or find useful, and what the industry
goes for may in the end be quite unscientific. Actually, if we ever
want to find out about the nature of such relationships, we would have
to establish independently what is objective knowledge (and there is
nothing like falsifiability for this purpose), what people do (and eve
n there usefulness is far too general and too subjective), and then
see whether there is a correlation. Once we would know about that, we
could then see what motivates what (for which we would again need
testable claims).

Another linguist who conflates quid juris and quid facti and in this
way follows (and even refers to) Feyerabend is John Goldsmith (1998).
For those interested in what Goldsmith said and what my views on that
are, cf. Ploch (2001), or send me an email
(plochlinguistics.wits.ac.za).

It would in my opinion help the discussion if those who support the
kind of anti-Popperian metatheoretical notions presented and argued
for by Feyerabend, Goldsmith and Everett looked at Popper's (1983:
xxvi-xxx) "List of Examples Chosen Almost at Random", where Popper
points to a number of "interesting cases in which refutations led to
revolutionary theoretical reconstructions" (p. xxv) and talked about
falsifiability (as Popper referred to it, falsifiability1) in this
context. Be that as it may, I would like to state that, since what
people call science and what constitutes truth are two altogether
different objects, which are in no a priori relationship, it can be
said that even if there was not a single case in the history of any
scientific discipline where refutation led to some new insight, this
would still not change the inherent subjectivism of concepts like
usefulness, on the one hand, and the logical validity of
falsifiability, on the other. Actually, anyone interested in finding
an explanation (not a subjective description useful to them) of what
scientists did find useful and why (Kuhnian) paradigms change, there
would be nothing they could do other than make testable hypotheses
about the matter and try ingeniously to destroy them. In this way,
they could gain some objective insights about usefulness (but no safe
knowledge, which is always impossible).

Another point I would like to make is that I find Everett calling
falsifiability a mantra out of place. In the end, it is the
subjectivists, relativists, sociologists of knowledge, historicists,
instrumentalists, etc., i.e., all those who think they can do without
testability, who have no formal and objective way of saying something
about truth and knowledge. They still make claims though. As Popper
has shown, this is the way of astrology, Freud's psychoanalysis,
Adlerian individual psychology, and I would like to add the Phonetic
Hypothesis, i.e., the claim that phonology is motivated by properties
of some phonetically defined system, the Grammaticality Hypothesis,
the Optimality formalism when taken to be a scientific theory, amongst
others (cf. Ploch, in prep., in print1, in print2). Note: ranking
systems can certainly be useful (and I see no problem with making such
a claim), if you decide a priori what the constraints you WANT are;
this way, you get the ideal output, given what you WANT. The only
thing of course is that it is impossible to provide evidence for a
violable constraint (this is a matter of logic, cf. Ploch, in prep.),
simply because a violable constraint is not testable (only the system
of constraints may be testable, which does not help, cf. Ploch, in
prep.). So in relation to Everett's implied accusation that I repeat a
mantra when I defend falsifiability, I would like to state that I
refuse to believe in anything PRECISELY BECAUSE I can instead, without
there any belief being required on my part, refer to Popper's
falsifiability as an objective and objectively arguable criterion of
demarcation between science and pseudoscience, on the one hand, and a
great deal more sophistication and even a whole epistemological basis
that Popper has worked on for around 60 years.

Furthermore, Dan Everett's wish to 'pick a theory' or to explain why
theories get maintained in the face of apparent counterexamples in
combination with his claim that falsifiability is not useful for that
is correct (in some way) but no argument against falsifiability.
Indeed, falsifiability does not in some exclusively logical way pick a
theory (by 'picking' I mean the action of making a decision to
participate in a certain theoretical enterprise) nor does it explain
why people do what they do (as pointed out above), e.g., why they
stick to an obviously falsified theory. There is however no problem in
that, that falsifiability would have to answer for. To explain this,
let me first say something about 'picking' a theory, and then about
why people stick to falsified theories.

(1) There is never any need to have a logical testable reason to pick
a specific theory . At any given moment in time, I know about certain
theories, and to the extent to which I am aware of their predictions
and to whether these predictions are testable I can say in relation to
any pair of them, which one is closer to the truth. This does not mean
that I therefore have to or should 'pick' that theory in some
psychologically committed or in other way exclusive sense. Simply do
not pick (in those senses). Do (in a testability-based manner) what
interests you. This means, practically and as indicated above, that if
a theory is not testable (=falsifiable1, not falsifiable2), change it,
or try to find a test, or if it is refuted, 'drop' it, or change it,
add an auxiliary hypothesis (for which you will have to find
independent evidence, if it is supposed to be scientific, not merely
useful), do whatever you WANT (which is not derivable from
testability!), and testability will tell you at all times, as long as
you stay testable, how well your theory is doing in relation to other
theories. That is to say, if you want to continue to 'pick' a theory
that has some major problems, fine (providing you are not denying the
problems or are explicitly allowing for them by not applying the
hypothesis that would otherwise be proven wrong, or by using a
'saving' conventionalist stratagem, i.e., providing you are trying to
overcome the problems, and note, there are, as described, several ways
to do that).

Also, you can entertain more than one theory at a time. It is the
subjective or social commitment to particular theories or some theory
on the part of some which makes them 'pick' in the 'committed' sense,
and it is the conflation of quid juris and quid facti which results in
them attacking falsifiability because it does, supposedly, not explain
what people do pick. One of the great advantages of a scientific,
i.e., falsifiability-based, approach is that one need not 'pick' a
theory in any of these senses, but can simply investigate something
that one finds interesting (i.e., one picks in some other sense), and
one can do so while constantly checking what the truth and falsity
status of one's theory/theories is. There is no problem with that
whatsoever. In this way, how well a theory (even a small subtheory on
a particular problem) is doing may change all the time, or quite
rarely. This depends on the kinds of tests one comes up with and on
what nature is actually like.

(2) The question why people stick to obviously false theories is an
interesting one (this is not a testable statement, this is
subjective). There are in my opinion two main reasons why people
behave in such a manner. There can be, first, a methodologically sound
reason to stick to a wrong theory: All (non-trivial) theories are
wrong somewhere, and how well a theory is doing can change dependent
on the tests someone comes up with (or the money, or the technological
possibilities, etc.). So, as I said above, one can try to save a wrong
theory in a scientific manner: one can change certain details (or even
all details), or one can try to come up with good auxiliary
hypotheses. Of course, what you prefer is dependent on all sorts of
science-independent notions (e.g., your personal interests) because of
which falsifiability, which is NOT about what people do and why, need
not explain which theory gets picked and 'stuck to' and by whom.
However, given that somebody finds it USEFUL to use falsifiability, we
can make predictions about what kinds of theories they will avoid
(e.g., those established on James' usefulness criterion). The other
main reason why people stick to false theories has to do with people
being psychologically attached to theories. Let me repeat that
falsifiability is about a logical relationship between strictly
universal statements and predicted and excluded existential
statements. Why people become 'attached' is an interesting question,
but falsifiability (a logical relation) has nothing to say about it.
However, if you want to investigate why people become addicted, you
would again have to make testable/falsifiable claims, and again,
falsifiability would be USEFUL. So if anyone is interested in why
scientists stick to obviously falsified theories, there is little
point in rejecting falsifiability for not preventing people from doing
so, but it would make much more sense to have a look at the psychology
of research (cf. Kuhn, who is much more in agreement with Popper than
is usually let on by those who try to use Kuhn against Popper, cf.
Schilpp 1974, but also Popper's counterarguments, ibid.) and the
psychology of academia (cf. Ploch 1999).

There is a further case where Everett's representation of Popperian
philosophy is inadequate. Reading Everett's description of Lakatos'
views, one could come to the conclusion that Lakatos' was quite in
opposition to Popper's views. For example, Everett wrote this:

"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."

Neglecting that I find Everett's implication that Popperian
falsifiability is 'pompous' quite out of place, Lakatos has not argued
against Popper's view 'so strongly'. What Lakatos actually says in his
main paper on the subject "Falsification and the methodology of
scientific research programmes" is this:

"Its [Lakatos' scientific enterprise's] main aspects were developed
from Popper's ideas and, in particular, from his ban on
'conventionalist', that is, content-decreasing, stratagems. The main
difference from Popper's original version is, I think, that in my
conception criticism does not -- and must not -- kill as fast as
Popper imagined." (Lakatos 1970: 179)

This statement sounds far less anti-Popperian than Everett's stance.
It should be mentioned here however that Lakatos misinterprets Popper
here too to some extent (but not to the same extent as Everett does):
Popper's intention was not to kill theories in the sense that if a
theory is wrong we have to propose a completely different one. Popper
had no problem with auxiliary hypotheses as long as they did not
immunise some hypothesis against refutation, or with changing certain
details of a theory. How much effort you want to spend on saving a
theory that is wrong somewhere is not derivable from Popper's
falsifiability nor did Popper ever say that it is (as far as I know;
if he did, I would like to know, but even if he did, which would
surprise me, then this would not be an argument against falsifiability
but possibly against Popper's stance on how quickly one should dismiss
a theory, which, as I have already stated, is not derivable from
Popper's falsifiability).

In may well be useful in this context to also consider the following
statement by Lakatos (from the same paper, p. 178-179):

"Now some of those who had already learned of the collapse of
justificationist rationality now learned, mostly by hearsay, of
Popper's colourful slogans which suggested naive falsificationism.
Finding them untenable, they identified the collapse of naive
falsificationism with the end of rationality itself. ... // But Kuhn
overlooked Popper's sophisticated falsificationism and the research
programme he initiated. Popper replaced the central problem of
classical rationality, *the old problem of foundations*, with *the new
problem of fallible-critical growth*, and started to elaborate
objective standards of this growth. In this paper I have tried to
develop his programme a step further."

Apparently, Lakatos was not at all 'so strongly' against Popper as
Everett lets on. Lakatos was actually of the opinion that Popper's
falsificationism was sophisticated (cf. the last quote), not
"pompous", as Everett thinks. This also means that Roger Lass too has
a view of Popper that is, to some extent, based on anti-Popperian
'hearsay' (to use Lakatos' term from the quote above) when he states
in his contribution to this discussion: "If simple-minded Popperian
falsifiability were the only kind, then all science would be subject
to the strictures that Lakatos used against Popper decades ago." The
simple-minded Popper Lass has in mind is not at all that
simple-minded. Even Lakatos admits that there is a great deal of
sophistication in Popper's falsificationism. As will become clearer
below though, Lakatos misrepresents Popper too: there is simply no
Popper that mixes what Lakatos without any evidence calls 'naive'
falsificationism and sophisticated falsificationism.

And yet again, there is a relevant quote from Lakatos (1970), this
time from the appendix (p. 181):

"In an earlier paper, I distinguished three Poppers: *Popper*0,
*Popper*1 and *Popper*2. *Popper*0 is the dogmatic falsificationist
who never published a word: he was invented -- and 'criticized' --
first by Ayer and then by many others. [There is a footnote with
details that I omit here, SP] This paper will, I hope, finally kill
this ghost. Popper1 is the naive falsificationist, Popper2 the
sophisticated falsificationist. The *real* Popper developed from
dogmatic to a naive version of methodological falsificationism in the
twenties; he arrived at the '*acceptance rules' of sophisticated
falsificationism* in the fifties. The transition was marked by his
adding to the original requirement of testability the 'second'
requirement of 'independent testability', and then the 'third'
requirement that some of these independent test should result in
corroborations. But the real Popper never abandoned his earlier
(naive) *falsification rules*. He has demanded, until this day, that
'*criteria of refutation* have to be laid down beforehand... Thus the
real Popper consists of Popper1 together with some elements of
Popper2'."

Obviously, Lakatos is not in complete agreement with Popper, but
Everett's presentation of Lakatos is quite different (apart from that,
it is subjectivist, and Lakatos' is not at all, but like Popper,
objectivist and falsificationist).

I will come back to Lakatos' view because I think that he is mistaken
when he claims that the real Popper is a mixture of what he calls
Popper1 and Popper2; I will even quote Popper referring to this very
quote by Lakatos. Be that as it may (for the moment), the following
quote by Lakatos, which I include at this place because it is from the
same paper by Lakatos (p. 179-180) actually describes quite well the
non-Popperian conflation of quid juris and quid facti (or rather, the
subjective reasons for quid facti), as I have discussed it above, and
as we have seen it, on the Linguist List, in arguments against Popper
which were constructed on the basis of saying that his falsifiability
criterion does not explain what theories get picked by scientists:

"Kuhn certainly showed that the psychology of science can reveal
important and, indeed, sad truths. But the psychology of science is
not autonomous; for *the -- rationally constructed -- growth of
science takes place // essentially in the world of ideas, in Plato's
and Popper's 'third world'* [I do not agree with the identification of
Popper's third world with Plato's, SP], in the world of articulated
knowledge which is independent of knowing subjects. *Popper's research
programme* aims at a description of this objective scientific
*growth*. *Kuhn's research programme* seems to aim at a description of
*change* in the ('normal') scientific mind... But the mirror-image of
the third world in the mind of the individual -- even in the mind of
the 'normal' -- scientists is usually a caricature of the original;
AND TO DESCRIBE THIS CARICATURE WITHOUT RELATING IT TO THE THIRD-WORLD
ORIGINAL [i.e., what Popper was interested in, cf. same quote above]
MIGHT WELL RESULT IN A CARICATURE OF A CARICATURE. ONE CANNOT
UNDERSTAND THE HISTORY OF SCIENCE WITHOUT TAKING INTO ACCOUNT THE
INTERACTION OF THE THREE WORLDS [my emphasis, SP]."

We see again that Lakatos is not by far so supportive of Everett's
anti-Popperian view as is implied in some of Everett's contributions.
So, Everett's 'supporter' Lakatos did not at all conflate quid juris
and quid facti the way Everett does.

Since Everett also tried to make much of Lakatos treatment of
anomalies, let me put that in context too:

"But since unexplained anomalies always abound, I allow such
formulations [presumably, in opposition to Popper, SP]." (Lakatos
1970: 182)

So in Lakatos' opinion, Popper and Lakatos differ (if they do -- it is
not actually as settled as Lakatos makes it sound) with respect to
whether exceptions are allowed. I find that it makes little
difference: as long as one does not ignore the anomalies and does not
try to verify one's (strictly) universal statements, there is no
problem; then one can call the anomalies of one's theories
'exceptions', consider them 'allowed' (i.e., their existence does not
prevent one from trying to eliminate the anomalies in question by
using testability as a constant checking tool in relation to how well
one's theory is doing). It does not follow from Popper's demand that
criteria of refutation have to be laid down beforehand (which does not
even have to be seen as a temporal but could be viewed as a
logico-deductive restriction) that he claims that a theory that
results in anomalies must be dropped 'totally' (which is the
implication one gets from the misrepresentations of Popper's views).
Popper's demand is that you admit, if you find anomalies, that your
theory, as it stands, is refuted (summa summarum, so to speak). Popper
does no claim that you must therefore abandon (=drop) your refuted
theory (in the sense, that you must now do something 'completely'
different); you may as well try to change your theory inherently or
add something to it that decreases the number of anomalies. There is
nothing in Popperian methodology that would prevent you from doing so.
So Lakatos is also in some sense fighting a strawman-Popper, but to a
much smaller extent that Everett. Apart from that, Lakatos does not
doubt the usefulness of falsificationism at all, and is not supportive
of Everett's usefulness nor of William-Jamesian pragmatism (cf.
Lakatos 1970: 99n1).

The following quote from Popper should make clear that Popper did not
support the view that a refuted theory must be dropped (as opposed to
changing it or as opposed to adding a non-ad-hoc auxiliary
hypothesis):

"Yet on the other hand, some of the most interesting and most
admirable theories ever conceived were refuted at the very first test.
And why not? The most promising theory may fail if it makes
predictions of a new kind... Even Newton's theory was in the end
refuted; and indeed, we hope that we shall in this way succeed in
refuting, and IMPROVING [my emphasis, SP] upon, every new theory. And
if it is refuted in the end, why not in the beginning? One might well
say that it is merely a historical accident if a theory if refuted
after six months rather than after six years, or six hundred years...
// Even if a new theory ... should meet an early death, it should not
be forgotten; ... for bequeathing to us new and perhaps still
unexplained experimental facts, and with them, new problems..."

Popper does not imply that when certain facts about Mercury became
known (which proved Newton wrong) one had to drop Newton but merely
that the crucial experiment in question pointed to a fascinating new
problem, i.e., that Newton had a problem, i.e., that Newton needed to
be improved upon, a problem which was ultimately solved by Einstein
(for now, i.e, until a new test results in some new falsification and
in this way some new fascinating problem).

This is not the Popper presented to us by Everett (or even Lakatos).
On the contrary, Popper's methodology can be summed up like this: "If
a proposed solution is refuted through our criticism we propose
another solution" (Popper [1984] 1996: 67), and it is here that we
find that his statement can be misunderstood: Importantly though,
Popper does NOT say that this 'other solution' has to be so 'other'
that one can consider the original solution 'dropped' totally, or in
its 'essence' (which would be what Popper referred to again and again
as an essentialist mistake); this kind of distinction actually makes
no sense; it suffices if the new solution is 'other'/different and
gets less things wrong, i.e., is an improvement (i.e., it can still
have anomalies, 'exceptions', falsifying data, whatever we want to
call them). And if one cannot come up with such an improved version
now, Popper is not implying in his work that one should not or cannot
now continue to try to come up with one. As pointed out above, if we
change a certain detail of a theory or even large parts of the
original solution, we get anOTHER solution, if we add to the solution
an auxiliary hypothesis we get anOTHER solution. To accuse Popper of
having proposed that any refuted theory must be 'dropped' in its
totality is as illustrated non-sensical; Popper merely says that it
must be changed (i.e., it should be considered dropped, or, one to be
dropped, an 'eliminatura', as it stands -- in that one knows for sure
that one needs an alternative), and, as I have shown, there are
several ways to do it, all of them are most practical, i.e., useful.
The only limitations one has are those that we humans can apparently
not do much about (invoking 'usefulness' will not help either): we may
be blocked, we may not come up with a good idea, there may be too much
admin, not enough money, etc.

That this interpretation of Popper is correct can also be seen by
looking at something that Popper pointed out again and again (e.g., in
1973: 16), i.e., that "Einstein's theory contradicts Newton's, which
it likewise explains, and contains as an approximation". So Einstein's
theory contains Newton's theory as an approximation; this is Popper's
favourite example of where one theory (Einstein's) was a better
('other') solution than Newton's (which was refuted and 'dropped' in
some sense). So the refutation and 'dropping' of some theory (say,
Newton's) is not meant by Popper to mean that one has to do something
'completely different' (whatever that is supposed to mean; actually,
it is not a relevant notion). One just needs a new solution; this new
solution will typically contain the earlier (refuted) theory as an
approximation. Now what is pompous and questionable about that
(Everett's terminology)?

All in all, it should have become apparent that Popper, when referring
to falsifiability, was talking about a logical relationship and, when
referring to the method of trial-and-error, about a scientific method
based on falsifiability that allows the scientist to evaluate what
alternative solutions are worth, scientifically speaking. Other than
that, I have also shown that Everett misrepresents both Popper and
Lakatos quite substantially, that he attacks strawman-Poppers, and
conflates quid juris and quid facti. In addition, Everett supports the
notion 'usefulness' of the subjectivist William James, without ever
looking at what Popper had to say about James' or similar
(instrumentalist or near-instrumentalist, pragmatist, subjectivist or
relativist) views, or without arguing against Popper in the much more
appropriate way the taking seriously of Popper's views on
instrumentalism, subjectivism and relativism would have made possible.

Finally, I have shown that Everett's claims that "it [falsifiability]
makes no difference to practice, ever", or that "[t]here is a lot of
darkness in the world. Falsifiability sheds no light to walk by" are
ill-conceived (because he conflates practice with a logical
relationship) and unfounded (the sheds-no-light statement). To those
interested in truth and objective scientific growth, it appears to me,
falsifiability is the only light they have, and I am glad that Popper
spent a life-time on investigating methodological and epistemological
questions that in some way or other follow from the method of
trial-and-error.

By the way, if anyone is interested in why people and/or scientists
ignore trial-and-error (not all the time, of course), I would like to
point them to my (1999) paper.

Apart from that, it was mentioned by Lass and Larsen that simplicity
(Lass) or economy (Larsen) (or Occam's Razor, for that matter) can be
used as scientific criterion. There is not enough space here to
illustrate in detail why I consider this view either unnecessary or
mistaken. (Let me add that I otherwise appreciate Lass' and Larsen's
generally more supportive view of falsifiability.) In short, there are
two types of simplicity that are worth distinguishing, one is keeping
the number of assumed universal statements (which are logically always
unverifiable) low, the other one is keeping the number of predicted
existential yet unverifiable statements low. In my paper
"Metatheoretical problems in phonology with Occam's Razor and
non-ad-hoc-ness" (Ploch, in prep.), I explain why due to the logic of
falsifiability (a) the first type (keeping the number of assumed
universal statements low) is unscientific if it results in an increase
of predicted existential yet unverifiable statements and why (b) the
scientific kind of simplicity (the second type, keeping the number of
predicted existential yet unverifiable statements low) need not be
added to one's metatheory but falls out from falsifiability. (For
details, cf. my paper, in prep. Email me, if you are interested.)

- ----------


References:

Feyerabend, Paul K. (1975)
Against Method. [1978] edition. London: Verso.

Feyerabend, Paul K. (1987)
Farewell to Reason. London: Verso.

Goldsmith, John A. (1998)
Subject: falsifiability (was: autosegmental phonology), sender:
cog�ling-relayucsd.edu, to: coglingucsd.edu, May 1998.

Lakatos, Imre (1970)
Falsification and the methodology of scientific research programmes.
In: Lakatos, Imre and Alan Musgrave (eds.). Criticism and the Growth
of Knowledge. 3rd impression [1974]. London/New York: Cambridge
University Press. 91-196.

Ploch, Stefan (1999)
Egoist arousal of empirical ideas for self-enjoyment. In: Scott,
Gary-John, Evelynne Ki-Mei Mui and Hyun-Joo Lee (eds.). SOAS Working
Papers in Linguistics, volume 9. School of Oriental and Afri�can
Studies (University of London). 319-356.

Ploch, Stefan (2001)
Metatheoretical problems in phonology with the common criteria
simplicity (Occam's Razor) and non-arbitrariness (non-ad-hoc-ness).
In: Lee, Hyun-Joo and Sam Hellmuth (eds.). SOAS Working Papers in
Linguistics, volume 11. School of Ori�ental and African Studies
(University of London). 123-153.

Ploch, Stefan (in prep.)
Metatheoretical problems in phonology with Occam's Razor and
non-ad-hoc-ness. To appear in: Ploch, Stefan (ed.). Living On the
Edge. Phonological Essays. (Studies in Generative Grammar.) Berlin/New
York: Mouton.

Ploch, Stefan (in print.1)
Can "phonological" nasality be derived from phonetic nasality? Paper
presented at the 4th HIL Phonology Conference (HILP4), Holland
Institute of Generative Linguistics, Leiden. To appear in the
proceedings; edited by Jeroen van de Weijer, Vincent van Heuven, and
Harry van der Hulst. Amsterdam/Philadelphia: John Benjamins.

Ploch, Stefan (in print.2)
Link Phonology: a functional explanation of non-monotonicity in
phonology. Paper presented at the 37th Conference of the Chicago
Linguistics Society, Chicago, 2001. To appear in the selected
proceedings (2002).

Popper, Karl R. (1972)
Conjectures and Refutations: the Growth of Scientific Knowledge. 4th
edition. London/Henley: Routledge & Kegan Paul. First published
[1963].

Popper, Karl R. (1973)
Objective Knowledge: an Evolutionary Approach. Reprint (with
cor�rections). Oxford: Oxford at the Clarendon Press. First published
[1972].

Popper, Karl R. (1983)
Realism and the Aim of Science. From the "Postscript to the Logic of
Scientific Discovery", volume 1, edited by W.W. Bart�ley, III. New
edition. London/New York: Routledge. First copy�righted [1956].

Popper, Karl R. (1996)
The logic of the social sciences. In: In Search of a Better World.
Lectures and Essays From Thirty Years, 64-81. First copyrighted [1984]

Schilpp, Paul Arthur (ed.) (1974)
The Philosophy of Karl Popper. Book II. (The Library of Living
Philosophers XIV.2.) La Salle, IL: Open Court.



Dr Stefan Ploch
Senior Lecturer
Linguistics (SLLS)
School of Literature and Language Studies
University of the Witwatersrand, Johannesburg
webpage: languages.wits.ac.za/~stefan

- ---------------------------------------------------
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Reply to Ploch

Date: Sun, 19 May 2002 11:31:15 -0300
From: Dan Everett <dan_everettsil.org>
Subject: Reply to Ploch


Professor Stefan Ploch's response is full of examples of how I have
misunderstood Popper, Lakatos, and so on. I have no doubt that I have
misunderstood a lot nor do I doubt that Professor Ploch has himself
misunderstood many of my claims. I could write a reply that dealt with
all the purported misunderstandings this way or that way, but I think
that this would be an unjustifiable expenditure of electrons and
clogging of Linguistlist's reader's mailboxes. So I will take a
different tack.

Let me preface my remarks by saying that I do not think that I called
falsifiability a mantra. Rather, I said that the citation of it among
linguists is mantra-like. In any case, that was my intention. 

The real issue at stake in these discussions is 'Objective Truth'. This
is what exercises Ploch, Bouchard, Odden, and others. My ex-colleague at
the University of Pittsburgh, John McDowell, has an interesting article
'Towards rehabilitating objectivity' in which he argues against James,
Rorty, and others quite cogently (not, however, effectively in my
opinion). He paraphrases James as saying "... Truth is not something we
have to respect, but a pointless nominalization of the useful adjective
we apply to beliefs that are getting us what we want." I go with James
on this one. Or as Rorty says (both of these articles are found in the
Blackwell book, _Rorty and his critics_, edited by Robert Brandom,
2000), "Worry about convincing your peers and truth and the world will
take care of themselves". Likewise, I find myself in agreement with
Rorty on this, as on many other issues. I believe that 'objectivity' is
a platonic residue of the notion that we can come closer to
understanding 'reality'. I don't believe this. I also believe that
'truth' is a compliment paid to a well-justified argument (also
paraphrasing Rorty). 

Continuing with this strand for a moment, the strongest philosophical
forces in linguistics are arguably platonic conceptions of Truth and
Cartesian rationalism, both of which I believe to be unhelpful and
fundamentally counterproductive. 'Plato's Problem' for me, to take an
example, is just this: "Why would anyone ever think Plato was even
relevant to modern inquiry on languages"? 

Falsifiability, as Dr. Ploch makes quite clear, is only intelligible as
part of a set of assumptions which includes 'reality', 'objectivity',
and 'truth'. Gosh, I reject all of these notions. Therefore,
falsifiability has no resonance with me. If I reject these notions,
however, do I believe in the growth of knowledge, the advance of
science, the legitimacy of intellectual effort? Well, yes, as a matter
of fact, I do - in each of these cases. To me, the advance of a science,
say, linguistics, occurs when its practioners agree that it does. A
hypothesis is useful to the degree that more and more practioners find
it more and more justified. To justify a hypothesis or a field or a
political position to me is to include more and more people and more and
more facts in better and better justified propositions. Falsifiability
is simply unavailable to me in this enterprise. 

Consider the Optimality Theory practioner who believes as I do vs. one
who believes in 'reality', 'objectivity', and 'truth' (and, once again,
I find myself in agreement with Rorty when he refers to these notions as
the modern man's religion). Let's call the former the NTOT position (No
Truth Optimality Theory) and the latter OTOT (Objective Truth Optimality
Theory). For the NTOT practioner, the theory can be useful just in case
it is clear, makes one think about things in ways they have not thought
about them before, causes one to look where they have not looked
before, and so on. However, I do not believe that there is any substance
to the question as to whether OT constraints are 'discovered' or
'invented'. On the other hand, an OTOT researcher will think that there
is a difference. OTOT constraints are part of a universal set hardwired
into the brain and they can only be discovered, never invented. But I
must ask myself, does this latter position make a difference to the
practice of phonology? I cannot see how. (I can think of some immediate
responses readers might have, but won't deal with them here.) 

So the issues are much deeper than they might have appeared in the
course of the linguistlist discussion. Ploch brings this out well. Let
me close off this discussion by urging my colleagues to take a book on
pragmatism to the beach or their mountain cabin the next time they have
a vacation and have some fun (of course, you might prefer to read some
Jack Kerouac, which is just as good). Read about another way of
approaching the world - as justifying our efforts to our fellow humans,
rather than finding some extracultural, superhuman device that separates
the right from the wrong or the good guys from the bad guys, or (as is
far more common in faculty discussions of their graduate students than
most graduate students ever dream) the smart people from the dumb. 

In any case, the next time you read an article that urges
falsifiability, ask yourself why it does this. Is it making a
well-justified case for its claims and are these claims useful to your
goals? Fine, then falsifiability does no work. Is its case
poorly-justified and useless to your goals? Then falsifiability again
does no work. But above all, don't take things too seriously. Realize
you're only justifying your work to your colleagues. You are not
approaching the Platonic Logos. No need for robes as you work.

Most importantly, though, as the bar-tender in 'Three Amigos' says so
eloquently 'We don't want no troubles!' Peace.

Dan Everett

P.S. At a party once, when some of these issues were being discussed
(not by me, I was just a grad student), John Searle went to his
refrigerator and got me a beer. He said, 'Dan, I just invented a beer
for you'. Pretty good argument if you want to go that way. 
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue