earman - carnap, kuhn, and the philosophy of scientific methodology
TRANSCRIPT
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
1/29
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
2/29
Carnap, Kuhn,and thePhilosophyof
Scientific Methodology
John
Earman
/
1
Introduction
For the past tw o decades logical positivismhas served as a w h i p p i n g
boy.
B y
e m p h a s i z i n g th e shor tcomings
of
this
failed
philosophical,
pr og r am, the v i r tues of the new
postpositiyist
pTi|osop_hy__of_sciencc
arc m a d e to seem more lustrous. It is, of course, n ot su r p r i s i ng to
f ind
such po lemica ldevices emp loyed , s ince they
are
c o m m o n
to
th e
rhetor ic of revo lu t ions ,
w h e th e r political, scientific,
or philosophical .
Or so the s t anda r d assessment wou ld go . W h a t I find askew in this
assessment is the not ion tha t a philosophical revolution asopposed
to
an
evo lu t i on
has
taken place. For a l though
I am no apologist for
logical posi t iv ism,
it
doessccoiJojme t h a t m a n y
ofthe t h e m e s f jJ T
so-called
postpositivist philosophyofjscicnccarcextensionsof ideas
f o u n d in the
wr i t i ngsoLCarnap
an d
other leading logical positivists
an d logicalemplficlstC^But m ypurposehere is not to
cont r ibute
to
a
revisionist history of philosophy. Rather , I aim to pay homage to
both Carnap
an d
K u h n
b y
noting some str iking similar i t ies
an d
also
some s t r ik in g di f ferences . Theses imilarities
an d
di f ferences
are
use-
ful
in helping to focus
some
of the
still
unresolved issues about the
n a t u r e of
scientif ic
methodology .
2 Logical Positivism, Logical Empiricism,,and
Kuhn's
Structure of Scientific
Revolutions
T he member s of
the
Vienna Circle often took votes on the issues
they
debated.
W h il e
I
have
n o
documen ta r y ev idence
to
of fer
of an
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
3/29
J o h n E a r m a n
ac lu a l
vole,
I am m o r a l l ycertain o f
what
the
rcsull
wo u ld
have been
fo r
ihc question,
Ver i f ica t ion
is a
relation between what
an d
w h a t ?
In x is verif ied by y, j&fs presumably a sentence. An d it istempting
to lake^jHo
be a
fact,stale
of
a f fa i r s ,
or
something
in
iheworldlhal
makes A C iruc
and can be
d i rec t ly ascertained.
But an
attempt
to
compare language
and
iheworld would have slruck iheCircle mem-
bers
as of a
piece w i th
the
metaphysics
thaiihe ver i f iab i l i ly principle
of meaning wassupposed lo
banish.Their a l t e r n a t i ve w as
lo
lake y
lo be another sentence, for then the relationjjctwecn x and yis an
.
unproblcmalic
logical
r c j a t i o n . _T h . 5
move, however,
seems
lo
leave
j
us in
ihc
same
metaphysical
th i cke l
since ver i f ical ion would seem
to
/
require thaiyis a true sentence, and is not a truesentenceone that /
corresponds to the facls?
The
escape lhat some
ofihe
circle members
soughlwas a
resort to
a
coherencetheory
of
truth. E v e n t u a ll y ,however, Carnap
abandoned
resort,
presumably
because
of a
combinalion
of
the
drawbacks
of
thecoherence account of truthand
the
allraclivenessof '
theory of truth. Bui w h a l I wish to emphasize here are ihe
qual i f i -
calions that Carnap put on any
talk
about comparing stalemenls
wi lh
facls. In
Truth
and Confirmation (1949) he
emphasized
lhat he
preferred lospeak of
confronling
slatementswilh facts:
There has
been
agood deal of dispute as to whether in ihc procedure
scientific
test ing
slatements must becompared
with
facts or as lo whe the r such
comparisons
be
unnecessary ,
if
nol impossible. If "comparison of s lalerncnl
wilh
fact"
m e a n s
ihc
p rocedure
which
we
called
ihc
firsl
opcralio^rylhcn
it
mus
be
adin i l lcd
lhal ihis
procedure
is nol
on ly
possible,
b ul
even indis-
pensable
for
scientific
testing. Y el
ilm u s
be remarked
lhal
ih c
fo rmula t ion
"comparison
of
s t a temen t
an d
fact"
is nol
unobjeclionable . First,
ihe
conccpl
"comparison" is not qu i l c appropr ia te here. T wo objecls can be compared
in regard lo a proper ly
which m
ir r j jarnr l n r j ~ r them in varia
We
ihcrefore
prefer lo speak of "cwnfronlalihH"
ra lhcr ihan^companso l
Confron ta t ion isun ders tood toconsX^jrjJiMnng out as to
fact
is
such
as isdescribed in
the
s ta tem ent , or, lo
express
it
differently ,
as
lo whe the r th e
s ta tement
is
true
to
fact.
(1949, 125)
Carnap then continued w i th a
passage
that might haveservedas an
advertisement for Kuhn's S t ruc t u re
o f
Scientific Revo lu t i ons .
Fur the rmore , ihc
formula l ion
in icrms of "comparison," in speak ing of
"facls" or
"realities"easily
Icmpts one inlo th e absolutislic
view
according lo
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
4/29
JJ
C a r n a p , K u l i n , an d
the
Philosophyof Scientific
Methodology
which
we
arc
said
to
search
for an
absolulc reality whose n a t u re
is
assumed
as
fixed i n d e p e n d e n t l y
of th el anguagechosen for itsdescr ipt ion .Thea n s w e r
to
aques t ionc o n c e r n i n g
reality
however
depends
n o t
on ly upon
that"reality"
or upon
th e
facts, but a lso upon the s t ructure (and the set
concepts)
of the
l a ng u a g e used for
the
descr ipt ion . In t rans la t ingone l a n gua ge i n t o a n o t he r
the
fac tua l
con ten t o f an empir ica l s ta tement c a n n o t a lways
be preserved
u n c h a n g e d .Such
c ha n ges
a rcinevitableifthe
s t ructures
o fthetw ol anguages
di f fe r
.inessential
po in t s . (1949, 125-126)
H e r e
we
havetw oof the key theses
oTthe^gostpositivisl
ph i losophyi
of science:, the
noncx i s t cnce
of ne u t r a l
facts
and
i n c o m m e n su r a b i l l y
in the foriTiof ' ' f tiftn^TrofiriKrrtTanshrtabHfly. NotcTttMrt-tucsc theses
were p ropou n ded i n the m id 1930s/for a l though I
have,
fo r thesake
ofconven ience ,quoted f rom the 1949 version o f
C a rnap ' s
paper , th e
re l evan t passageis
also
in
"W ahrhci l
u n d B e w a h r u n g "
(1936)./*
|
Given
these K u h n i a n themesor should wer a t h e r s ay Ca r n ap i an
th e me s? one
m i g ht predic t th a t Ca r n ap wou l d h av e f ou n dStruc ture
phi losophical ly congen i a l . That
this
w as
indeed the
case has been
d o c u m e n t e d b y
Rcisch
(1991).
St ruc tur e
was publ ished as par t of
the
I n t e r n a t i o n a l Encyc loped ia of Un i f i ed Science, of
which
Ca r nap w as
an
associateedi tor .A f t e r r ead ing the comple ted manusc r ip tfo r
Struc -
ture , Ca r nap wr o te to K u h n in A p r i lo f 1962.The text of
the
let ter
is reproduced in
Reisch
1991. I
will
quote f rom notes wr i t ten in
Carnap ' s a rchaics h o r th a n d .Carn ap beg ins wi th
a
piece
o fD a r w i n ia n
evolu t ion
an d
t hen adds,
c o n c e r n i n g
K u h n ' s
thesis,
In
a n a logy
to
this
one
has
to
un de r s t a n d t he de ve lopme n t
o f
scientific
the-
ories:
n ot
directed to an ideal[ true] theory, the
o n e
t rue theoryof the wo r ld ,
but
evolu t ion as a
step
to a
bel ter f o r m ,
by
selection
of
one
ou l of
several
compe t ing forms .The selection is
m a d e
on
ihc
basis of preference in
ihc
c o m m u n i t y
o f
scient is ts . M an y
factors,
sociological,
c u l t u r a l , . . . , a re involved.
Nol :
v
w e
a re a pp r oa c h in g i r u lh , bu t : we a rc i m p r o v i n gan instrument
An other im po r tan t though large ly tac it them e of S t ruc t u r e ,a
holist ic
view
of
m e a n i n g ,
c an
also
be
seen
as
emerging f rom
the
wr i t ings
of
the logical empir icists in the
1950s.
Hcmpe l , for example, took to
heart Quine 's at tack on the
analyt ic /synthet ic
d i s t i nc t ion ,which m ay
be taken to e m b o d y the thesis that there is asharp d is t inc t ion to b e
d r a w n be tween two f u nc t i on s o f
l anguage :
one to specify m e a n i n g ,
the other to make empir ica l asser t ions . Applying the mora l to sci-
ent i f ic
theories, one arr ives
a t
the conclus ion tha t there i s no p r inc i -
,*>
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
5/29
1 2
J o h n
E a r m a i i
pled
way to
d i s t inguish those postulates
of
th e
theory
tha t
m ay
proper ly
b e
d u b b e d m e a n i n g postulates.
It is
then
but a
seemingly
short
an d t e m p t i n g
step
to the
fur ther conclus ion tha t
all
the
pos-
tulatesof the
theory
function to specifythe mean ing of the constit-
u e n t
te rm s
a n d
thus tha t
an y
s ignif icant change
in
the theory
implies
a
change
in me an i ng .
This route
to
seman t ic hol ism cann ot
betraced to any of
Carnap 's
writings. Indeed, in his response to HempePs contr ibut ion to the
Schi lpp volume (1963), Carnap at tempted to use the not ion of the
Ramsey
sentence of the theory to
ident i fy
the postulates of the theory
tha t
"merely
represent
meaning relations"
(1963a,
965).
And in h is
last bo o k , Phi losophica l Foundat ions o f Physics
(1966), Carnap mai iv
ta ined th a t a sh a r p analyt ic-synthet ic dist inct ionis ofs u p r em e im-
portance for the phi losophy of science" (1966, 257). There
is,
however, another C arn apian route to sem an t ic hol ism, bu t tha t route
m u s t be
traced
all the w ay
back
to
Carnap ' s a t t e m p t
in the Aufbau
(1928) to expla in how scientific objectivity can
emerge
from a
recon-
struction
tha t
starts from
a
phcnom cna l i st ic
basis.
A n
explora t ion
of
this
m a t te r
w ou ld take
m e too far
afield;
I
willsim ply refer
the
reader
to Michael
F r i edman ' s (1987)
i l lumina t ing discussion.7
^ JV
/
3 Carnap's Relativism
losophy an d
Logical
S y n t a x " (1935)
proclainicd the
relat ivi ty
ol
a ll
philosophical
theses tol anguage .This
relativity
w as
supposed
tohold
the key to
so lv ing ,
or
rather dissolving, t rad i t ion al phi losophical dis-
putes .
Suppose,
for
example ,
on e
phi losopher
asserts,
" N u m b e rs
are -
primit iveentities,"
whi le
another proclaims, "Numbers are classes of
classes." They
m ay," C arn ap w ri tes, "philosophize w itho ut
en d
abou t
the
question
of
what numbers rea l ly are ,
but in
this
w ay
they
will
never
come
to an
agreement
(1935, 450).
If,
however, they
are
acute
en o ugh to
recognize Carnap's relativity principle, they
will
quickly
realize
that one is asser t ing, "In Language LI (Peano), numerical
expressions are elementary expressions," whi le the
other
ism a i n t a i n -
ing, "In
Language
L?
(Russell) numerical expressions
are
class
ex-
pressions of the secondorder.
"Now these
assertionsare
compat ible j
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
6/29
C a r n a p , K u l i n ,
an d
th e P h i l o s o p h y
o f Scienti f ic
M e t h o d o lo g y
wi th each o th e r an d bo th a rc t r u e ; the cont roversy has ceased to
exist"
(1935,451).T h is m o d e l for resolving phi losop hical dispu tes in
the
ph i losophy
of
m a t h em a t ic s
w as
supposed
by
C a r n a p
to be
cx tend-
iblc q u i te b road ly to phi losophical disputes, such as p h e nom e na l ism
versus
mater ia l ism an d the quest ion of whether space- t ime poin ts
have
an
ex i s tence i ndependen t
of
physical events .
Some, like Donald Davidson (1973), have found an air of paradox
in the
fact
tha t say ing i n one brea th (as Ca r nap th ou g h t he could)
tha t 5 is t rue in L \,
LS,
Las, . but false in L ? ,4, .. . seems to
presuppose
a
n e u t r al
m e ta f ra m e
wi thin
wh i chall thelanguage f rames
can
be
treated.
6
Others, l ike Michael Fr iedman (1992), have argued
t ha t
Carn ap ' s re la t iv ism is un de rcu t byGodel's incomple teness theo-
rems , wh ich show tha t
n o
such neu t r a l me ta f r ame
is
avai lable.
M y
objections are more localan d tactical.
Myfirstc o m p l a in tisth a t
Carnap
assumesw h a t
needs
to be proved .
Accord ing toCarn ap ' s "pr inc ip le of tolerance," we are free tochoose
whateve r
l a n g u a g esystem
w e
like.
T he
decision
islargelya
p r ag m a t ic
affair , t u r n i n gon such mat ters as efficiency an d
f ru i t fu lness
fo r the
-purposes a t h and . But to a p p l y the slogan o f "free
to
choose to
dissolve,
say,
the
p h e nome na l i sm v e r su s
m ater ial ism
deba te assumes
tha t a p h en om ena l i st ic langu age has been produ ced that shows how
physical object talk can be reduced to
talk
about sensa data or , as
C arn ap preferred, m om en tary total experiences ("Elementarer lcb-
nissc ).
This,
of course, is exactly what Carnap tr ied to do in the
Av/lau.
But by h is own admiss ion ,
th e
a t t em pt
has to coun ted as a
f a i lure
if, ash e o r ig ina l l y assumed , the
reduct ion
has to
proceed
v ia
explicit def in i t ions
an d
explicit Iranslalions. Similar ly,
to
a p p l y the
"free to choose slogan to dissolve disputes about the ontplogical
status
of space-t ime points assumes tha t i t can be shown how
space-
t ime poin ts can b econstructed out of events, such ascoincidences of
particles. Advocates
of
relat ional theories
of
space
an d
t ime repeat-
edly
claim that this can bedone an d even tha t i t has beendone. B ut
n o n e
o f
the c la ims s tands
up to
scrut iny.
7
I am
e mp h a t i c a l l y
n o t
c la iming tha t materia l ism
is
correct
or
that space-t ime points con-
strued as irreducible enti ties are essential to physics. Rather, I am
c la iming that the dissolution of the tradi t ional disputes on
these
matters is not as easy asCarnap made it seem.
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
7/29
M
John E a r m a n
A n o t h e r
c o m p l a i n t arises f rom
the b r ea t h t ak i ngscope of the in-
t ended appl ica t ion
of
Carnap's
di sso lving s tra tegy . A m on g
the
"phi lo-
sophical" disputes that C a r n a p proposed to treat in th i s way were
such
matters as
whether time
is finite or
inf ini te
an d
whether
the
world isdetermin is t ic .There isobvious lyavery s l ippery s lope here.
If the quest ionabout the fini lude of t ime is a philosop hical q uestion
in
the
re levantsense, then
why not the
q uest ion
of wh ether the
world
began from a b ig-bang s ingular i ty?An d if this latter question is a
phi losophicalqu est ion
in the
relevant sense, then
w h y n o t
otherdeep
ques t ions
in
cosmology?
B ut
more
impor tan t , one
does
n o t
have
to
go d o w n the slope to recognize the implausibihty of Carn ap 's pro-
cedure .E ven
i f one
agrees
to
talk
about
t ruth
in Lrather
than t ru th
period,
there is no plausibi l i ty to the idea that whether t ime is f in i te
an d whe ther determin ism holds a re mat ters to be settled in L by
adopt i ng l inguis t ic rules for
L
rather than by consul t ing the facts .
C a r n a p ,
n o t
surpr i s ingly , acknowledged
the
po in t . Speaking
of the
determin ism
issue, he said,
Theob jection m aypcriiaps be raised at this point that the form of physical
laws
d ep end s upon exper imenta l resu l ts
of
physicalinves t iga tion ,
an d
tha t
it
is not d e t e rm i ned by a merely theoretical syntactical
consideration.' This
assertion isqu i te r ight , b ut we
m u s t bear
in m i n d the fact
tha t
the empir ical
results
at which physicis ts arr ive by way of their laboratory exper imen t sby
no means
d ictate
theirchoice
between
th e
determinis t ic
an d the statistical'fo rm
of
laws.
The
fo rm
in
w h i c h
a law is
to be slated
has
tobedecided
by an act
of
voli t ion.
Thisdecision, it ist rue ,
d ep end s
u p o n the empir ical resul ts , but
n ot
logically,
on ly
pract ica l ly .
The
results of
the exper iments show
merely
i l l . i l
on e
mode
of fo rmu la t ion wou ld be
m o r e
sui table t h a n ano the r . (1935,
455)
These s en t imen t s
resonate with those Carnap expressed three de-
cades la ter in comment ing on
K u h n ' s St ruc ture .
B ut here the senti-
ment s
arc not to the point. T he issuei s not
whether ,
for D u h e m i a n
or
o ther reasons,
the resultso f
exper iments ,
say the
recen t E ins te in ,
Rosen, Podolsky, and
Bell
type of exper iments ,
fail
to dictate the
acceptance o f
i nde te rmin i s t ic
laws. Rather the issue is whether deter-
min i sm is a scientif ic claim to be argued over the way one argues
over otherdeepscien t i f ic c laims, n on e
of
w hich ever gets
definit ively
settled by the dictates of experimental evidence; o rw hether
deter-
min ism is a
c la im that
can be
made t rue
by
l i ngu i s t i c
fiat in
L\
Lr , ,
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
8/29
Ca rn a p , K u h n , an d the Philosophy
of
Scientific
Methodology
35, . . . and false by l inguis t ic fiat in L
2
, LT , 37 . . .. . an d then w e
j u s t
p a ysour m o n e yan d takeso ur choiceo f language . Aga inI
re fuse
to give a global
a n s w er
to th i s que ry an d favor instead a
tactical
response.All in dica t ion s are tha t the debate over theimp l ica tions for
deter i j i in ism of the
Bell
inequa l i t ies and the
Aspect exper iments
be longs
to
the f o r m e r r a t h e r than to the la t te r . Indeed , ind ica t ions
are
t h a t
i n any l anguage
sys tem adequate
for the
f o r m u l a t i o n
o f
theoriestha tsavethe exper imenta l lyverifiable q ua n tum s ta ti stics ,the
laws m u s t
b e
inde termini s t i c .
9
I
suspect
t h a t
Carnap ' s
relativism
began
b y hisbe i n g
impressed
b y
th e ach ievements o f Frege, Russell, an d others in the ph i losophy of
mathema t i c s
and was fu r the red by a misplaced zeal for extend ing his
m o d e l for resolving
ph i lo soph ica l d i spu te s
in this area
to
a
broad
area of ph i losophica lan d scient if ic questions. O f course, wha teve r
the origins of Carnap's relat ivism, it or
s om e th i n g
like it could per-
hapsb e promoted on the basis of his doct r ine tha t language-neut ra l
facts do not exist. I find it di f f icul t to assess this matter, since I do
not f ind in Carnap ' s wr i t i nga he lpfu l explana t ion o f
this
In
th e fol lowing section
I will
c o m m e n t
on the related
doc t r i ne
o f
Feyc ra bc n d ,
H a n s on ,
an d
Kuhn tha t observa t ion
is
theory laden .
In closing this section, Inote tha t Carn ap displayed a con sistency
on th e
ma t t e r
at
h a n d
n o t
the
consistency that
is the
hobgoblin
of
l i t t le m i n d s but the magn i f i c en t consistency of a grand v is ionary . In
his
c o n t r i bu t i o n t o
the
Sch i lpp vo lume fo r Carnap ,
H e r b e r t Fcigl
(1963) ske tched a mind-body ident i ty
theory
that he was later to
elaborate
in his
f a m o u s
essay
The
'M e n t a l ' an d
t h e T h ys i -
caF"(1958). ' Clearly, the politically
correct th ing for Carnap was to
endorse Fcigl 'sapproach. Ins tead , he wrote , "it seems preferable to
m e to formula te the ques t ion [of mind-body ident i ty] in the meta-
lan guag e, not as a
factual
ques t ion abou t the w or ld ,but as aques t ion
conce rn ing the choice of language . Al though w eprefer a d i f f e r e n t
language , we m us t adm it tha t a dua l is t ic
l anguage
can be con structed
an d
used
w i thou t coming in to con f li c t
with
cither
the
laws
of
logic
or
with
empi r i c a l l y known
facts"
(1963b,885-886).
4 Kuhn's Re lativism
K u h n resists being labeled a relativist. I use
the
labelhere tore fe r to
three doct r ines o f Struc ture: the theory ladenncss of observat ion , th e
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
9/29
]G
John
Earnian
n c o m n i c n s u r a b i l i t y
o f
theories,
an d (l ie d e n i a l t ha t
there
is a
l l i co ry -
indcpcndcnl
notion
of
truth.
Pa r t of w h a t w as
meant
by the
theory ladcnncss
o f observation is
embodied in the
thesis that
what we see depends
upon
what we
believe,athesis
open
tochallenge
(see
Fodor 1984). I am
concerned
rather with
therelated
thesis
of the nonexistence of aneutral
obser-
vat ion language
in w h i c h
different theories
can be
compared.
M y
response is once again tactical.
That
is, without trying to adjudicate
thegeneral merits of the
thesis,
I claim that things
aren't
so bad for
actual
historical
examples.
Even
for
cases
o f
major
scientif ic
revolu-
t ions, we
cari
find,
without having
to go too far
downward toward
some t h ing like foundations for knowledge, an observation basethat
is neu t ra l enough
fo r
pu rpos e s
a t
hand. A
nice
e x a m p l e is
provided
by
Allan Franklin
(1986,
110-113), who shows how to construct an
experiment that
is
theory-neutral enough between Newtonian
and
special-relativistic
mechanics to unambiguously
decide
between the
predictions
of
these
theories for elastic
collisions.
The two
theories
agree on the procedure for measuring the angle between the
velocity
vectorsof thescattered particles,and the two
theories predict
differ-
~
en t
angles.
A b
More
generally,
I
claim that
in the
physical sciences there
is in
pr inc iple
always available
a
neutral observation base
in
spatial coin-
cidences,such asdots on photographic plates, pointer positions on
dials,
and the like. If intcrsubjective agreement on such
matters
were
n ot
routine, then physical science
as we
k n o w
it
would
no t b e
possible.
1 reject, of course, the posilivistic attempt to reduce
p'hysics
tosuch
coincidences.
A n d I readily acknowledge that such coincidences b y
themselves
are
mute
witnesses in the
tribunal
forjudging theories.
B ut
w h a tis required to make these mute witnesses articulate is not a
Gcstaltexperiencebut a
constellation
of techniques,hypotheses,and
theories: techniquesof
data
analysis, hypotheses aboutthe operation
of measuring instruments,
and
auxiliarytheories thatsupportboot-
strap calculations of values for the relevant theoretical
parameters
t ha ttestthecompeting theories.But Iagain assert thatto the
extent
tha t
this process cannot be explicit ly articulated but relies on some
su igeneris
form
of perception,the
practice
is not
science. This
is not
tosay, however, thatthevulgar imageofscienceas ab l i nd lyimpartial
enterprise
is correct, for the articulation uncovers assumptions to
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
10/29
17
Carnap, K u l i n ,
a nd the
Phi losophy
o f
Scient i f ic Method o log y
which d i f f e r e n t
scientists may assign vc fy
different , degrees
of con-
f idence .
B u t
this
sense
in
which
di f ferent scient is ts can
(mis lcad ing ly )
be said to see di f f e ren t th ings when looking a t the same phenom -
enon
is one with which a
probabil istic
or
Bayesian
epistemology the
k i n d
of
epis temology w hich
the
la ter Carn ap came to
advoca tem ust
cope on a routine basis ,
even
incases fa r away f rom the boundar ies
of
scient if ic
revolut ions .
H ow
these di f ferences
are resolved is
pa r t
of
the
Bayes ian ana logue
of
K u h n ' s problem
of
com m un i ty dec is ion
on
theo ry choice. Kuhn 's problem
will
b e
encoun te red
in
the
follow-
in g section, an d the Bayesiananalogue will be discussed in sections
8 and 9.
T he m a t te r o f incom m ensu rab i l i ty i s m uch more diff icult to discuss
for tw o reasons.
First ,
it is tied to di f f icul t is sues about me an in gan d
reference tha t
I
cannot broach
here.
Second, issues about incom-
mensurab i l i t y
present amorphous and shi f t ing targets. In Struc ture ,
for
example , i ncommensurab i l i t y
was a
label
for the
entire constel-
lation
o f
factors that lead proponents
of
d i f f e ren t pa rad igms
to
talk
past on e another . In recen t years Kuhn has come a round to a
more
Carnap ian
or
l inguis t ic fo rmula t ion
in
which i ncommensurab i l i t y
is
equated
wi th
un t rans l a tab i l i t y . More
specifically, the
focus
has
shif ted
from paradigms to theories, and two theories are said
to
b ei n c o m -
mensurab le just in case
there
is no common language in to which
both can be fully translated ( K u h n1989, 10). I have no doubts about
K uh n ' s cl aims tha t
theories
o n d i f f e r en t
sides
o f ascientificrevolut ion
often
use
di f ferent " lexicons," that di f ferences
in
lexicons
can
m a k e
fo r a
k i n d
o f
un t r a n s l a t a b i l i t y ,
an d
tha t
in
t u r n this
exp la in s w hy
scientists r e ad i ng out-of-date texts of ten
e n c o u n t e r
passages tha t
make
no
sense
(1989,9). But I deny
thatthere
is
i n comm e nsu ra -
bi l i ty /untransla tabi l i ty
tha t makes
for
insuperable difficulties
for
con-
firmation or
theory choice
(a
phrase
I
don't l ike
for
reasons
to be
given below)
in the
s tandardly c i ted cases
of
scientific revolut ions
such as the t ran si t ion f rom N ew ton ian to special- rela t ivis tic m echan ics
an d
the
su bse qu e n t transi t ion togeneral rela tivi ty . N ew ton ian , spe-
cial-relativistic,
gcncral-relativistic, and
m a n y
other theoriescan all
be
f o rm u l a t ed
in a
common language,
the
language
of d i f fe ren t ia l
geometry
on a
four -d imen s iona l m an i fo ld ,
and the
cruciald i f ferences
in the
theories
l ie in the
dif ferences
in the
geometric object
fields
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
11/29
J8
John E a r m a n
postulated
an d the m a n n e r i n wh i c hthesef ieldsrelateto such
t h i ng s
as
particle orbits .
This
language is
an achron is t ic
and so may not be
th e best device
to
use when t r y i n g
to decide
various historical dis-
putes. But i t
does
seem to me to be an
app r op r i a t e vehicle
for
f r a m i n g an d an swe r i ng the sorts of quest ions of most concern to
w or k ing
physicis ts
an d philosophersof
science.
For
example,
on the
basis of the
avai lable
evidence, what is i t reasonable tob elieve about
the
s t ruc ture
o fspace an d t ime and the n a tu re o f gravi ta t ion?This
is not to saytha t the common language makes for an easy answ er. It
is
indeed a
d i f f icul t
business, but i t is a business that
involves
the
same sorts o f difficult ies a l ready
present
w h en test ing theories t ha t
lie on the same side of a scient i f ic revolut ion. Finally, so that there
can be n o m i s un de r s t a n d i n g ,
let
m e repeat: I am not c l ai m i n g ( h a t
w h a t
I
call
a
common l anguage prov ides wha t K u h n wan t s . It docs
n o t
show,
fo r
example , tha t
the
N e w t o n i a n
and the
E i n s t c i n i a n
can
b e
b rought i n to agreemen t about wha t
is and is not a
"m ea n i n g f u l "
quest ion
a b o u t
s im ul tane i ty .
But w ha t I do cla im is that
these
res idua l
e lement s
of i n c om m en s u r a b i li ty
d o no tu n d e r m i n es t anda rd accoun t s
of
theory test ing
an d
con f i rma t ion .
1 2
My response
to
worries
about the
appl icabi l i ty
of the
no t ion
of
t r u t h
tow hole theories iss im ilarly localan d
tactical.
In the Postscript
to the second edition
o f S t ru c tu re ,
Kuhn wr i t es ,
There
i s , I th in k , no
t h e o ry - i nd e p e n d e n t way to
recons t ruc t
phrases like 'reallythere'; the
not ion
of a
ma tch between
the
ontology
of a
theory
and its
' rea l '
c o u n t e r p a r t
in
n a t u r e
n ow
seems
to me
illusive
in pr inc ip le"
(1970,
206). I need n o t
d e m u r
if " theory" is
unders tood
in 'a very broad
sense to mean something l ike a conceptua l f r a m e w o r k som i n i m a l
tha t w i tho u tit "the world" would b e un d i f f e r en l i a t ed Ka n t i a nooze.
B ut
I do
d e m u r
if
theory
is
taken
in the
ordinary sense,
i.e., as
Newton ' s theory o r special - rela t ivi ty theory o r genera l - re la t iv i tythe-
ory.
13
For scient is ts are current ly working in a f rame in which they
can
say, correctly
I
t h i n k ,
tha t
the
match between
the
ontology
of
the
theory and i ts real counterpar t in n a t u r e is
better
for th e special
theory
o f
relativity
an d evenbetter for the
gen era l theory.
O f
course,
to get tothis position required tw om a jor coneptua l revolut ions . H ow
such revolutions affect theory choice, or as I would prefer to say,
theory testing an d con f i rma t ion , rema ins to be discussed.
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
12/29
]9
C a r n a p , K u h n ,an d the Ph i losophyo fScientif ic Methodology
5
Kuhn's
Accountof
Scientific Revolutions
Carnap,
as wehave
seen,
found
Kuhn's
Structure
congenial.
But
m a n y
philosophers
of the younger
generation,includingthose
who
prided
themselves on having
gone
beyondthe crudities of
logicalpositivism,
professed
shock and
dismay
at
Kuhn'saccount
of the
displacement
of
an old
paradigm
by a new
one.
For
those
readerswho do not
have
a copyofStructure to
hand,
hereare someof the
purplepassages:
Like
th e choice between competing poli t ical ins t i tut ions , that
between
com-
pet ing
parad igms
proves to
be a
choice
between
incompat ib le
modes
of
c o m m u n i t y
life.
.
..W h e n
parad igmsen ter, as
they
m us t , in toa deba te abou t
parad igm
choice, the i r role is
necessarily circular.
Each group
uses'
i ts own
parad igm to a rgue in t h a t p a r a d i g m ' s
de fense .
(P. 94)
As
in poli t ica l revolut ions , so in parad igm choicethere i s no s tandard
h ighe r t h a n th eassentof
the
re l evan t
c o m m u n i t y .
To discover how
scientific
revolu t ions arc
effected,
we
shall
therefore
have
to
e x a m i n e n o t on ly the
impact o f n a t u r e and logic, but also the t echn iques of pe rsuas ive a rg um en -
tation
w i th i n
th e
c jui lc
special
groups
tha t
cons t i tu te
th e
c o m m u n i t y
of
sci-
entis ts . (P. 94)
The
p r o p o n e n t s
of compet ing parad igms prac t ice the i r
trades
in d i f fe ren t
worlds. . . . Prac t ic ing in d i f ferent wor lds , the tw ogroups of scientists sec
d i f f e r en t t h i n g s w h e n they look from the same point in the
same
direct ion.
(P. 150)
In thesema t t e r sn e i t h e rproof n orerror is atissue.The
t r an s fe r
ofallegiance
from
pa r ad igm
to
parad igm
is a
con version
exper ience
tha t
c a n n o t
be
forced.
(P .151)
Before they can hope to c o m m u n i c a t e fully, on e group or the o ther mus t
exper ience th e
conversion
tha t we have been cal l ing a paradigm shif t . Jus t
because it is a
sh i f t between
incommensu rab les , the t rans i t ion
between
com-
pet ing
pa r ad igms
c a n n o t
be
m a d e
a
step
at a
time, forced
by
logic
an d
neu t r a l experience. Like a gcslalt switch, it must occur all at once
( though
n o t
necessarily at an
i n s t an t )
or not at al l . (P . 150)
M a n y
readers
saw in
these
passages
an
open
i n v i t a t i o n
to aralio-
na l i t y if not outright irrationality.Thus Imre
Lakatos
took
K u h n
to
besaying that theory choice is a matter of "mob psychology
(1970,
178), while DudleyShapcreread Kuhn assayingthatthe decision to
adopta newparadigm
cannot
bebased ongood reasons (1966, 67).
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
13/29
20
John
E a r man
K u l i n
inturn w asequ a l lyshocked b ysuch criticisms.In thePostscript
to thesecond edi t ion o f St ruc ture (1970),he professed
surprise
that
readers could have imposed such unintended interpretationson th e
above quoted passages.
I
will leave aside
the
u n f r u i t f u l question
of
w h e t h e ror not
K u h n
ought tohave
anticipated
such interpretations
an d
will concentrate instead on what, upon reflection, he i n t e nd e d
to
say.
K u h n ' s
ow n exp lana t ion in the
Postscript begins
wi th the
common-
place tha t debate over theory-choicecannot be cast in a form tha t
resembles
logical
o r
mathemat ica l
choice
(1970,
195).
B ut
he
has tens
to
add
that this commonplace
doesnot
imply that there
are no
good
reasons for
being
persuaded or that
these
reasonsare not
u l t im ate ly
decisive
for the group (1970, 195). T he reasons listed in the Post-
sc r ip t
areaccuracy,s implic i ty ,and f ru i t fu lness . The later paper Ob-
jectivi ty, Value Judgments,
and
Theory Choice (1977)
added two
f u r t h e r reasons:consistencyand scope.And asKuhn himselfnotes,
the final
list
does
not
d i f fer
(with
one
notable exception
to be
dis-
cussed later) from similar lists drawn from standard phi losophy-of-
scicnce texts (see also Kuhn 1983).
These soothing sentiments serve todeflate charges ofarationality
and
irrationality, but at the same time they also serve to raise the
ques t ion of how Kuhn's viewsare to be distinguished from the or-
thodoxy
that Structurewassupposed to upset.The answer givenin
the
Postscript contains
tw o
themes, which
are elaborated in Objec-
tivity."
First,
the
items
on
the
above
list
a re
said
to
function
as
values"
tha t can be
d i f f e ren t l y app l i ed ,
i n d iv idua l ly an d
collectively
b ym en
whoconcurinhonoringthem (1970, 199).Thus, thereis noneutral
algor i thmfor
theory choice,
no
systematic decision procedure which ,
properly applied, must lead each individualin thegroupto the s ame
decision.
Second, it
(supposedly) follows that
"it is the
c o m m u n i t y
of
specialists
rather than the individual members that makes the
effective
decision (1970,
200).
I
th ink
that Kuhniscorrectinlocating objectivityin thecommunity
of
sp ecialists,
atleastin the uncontroversialsensethat intersubjective
agreement among the relevant experts is a necessary condition for
objectivity. But how the communityof
experts
reaches a decision
when the individual members d i f fer on the application o f shared
values is a mys te ry that to m y mind is not
adequa te ly
resolved b y
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
14/29
21
C a r n a p , K u h n ,
an d ihc Phi losophy o f
Scicnlific Mclhodology
Struc ture or by s u b s e q u e n t wr i t ings . M ystrategy will be to
explore
these
an d related
issues from
the
perspective
o f
Carnap's
cpis tcmology.
6
Carnap
and
Kuhn Incommensurability?
The passage Reisch
(1991)
quotes from Carnap's letter toKuhnand
the passage from Carnap's shorthand notes I quoted in section 2
would seem to ind ica te
that Carnap
an d
K u h n were
in
subs tan t ia l
a g r e e m e n t
as
regards
paradigm
choice.
This
is
surely
the
case
w h e n
paradigm is
interpreted
to
m e a n
something l ikea l inguis t icframe-
w o r k . A t this
level
Carnap wouldagree, indeed, would insist , on th e
need tochoose, and hewould hold t h a t thechoice
is'a p r a g m a t i c
on e
whosedynamicsm ay well invo lve the
sorts
of factors emphas i zed in
K u h n ' s account. B ut w h e n
the
focus shifts to theories, as itdoes in
K u h n ' s later writings,
the
disagreement begins.
In the f i rs t
place,
K u h n ' s list
ofcriteria fortheory choice isconspicuousfor itsomission
of any
reference
to the
degrees
of
confirmation
or
probabilities
of
the
theories.
This is not an
oversight,
of
course,
but
derives both
from explicit doctrines, suchas thenonexistence of atheory-neutral
observation language, and the largely tacitbut pervasive anti-induc-
tivism
of
S t ruc t u re . Needless
to
say, this shunning
of
confirmation
theory
is
most u n - C a r n a p i a n .
B ut
even more anomalous from Car-
nap's
perspective
isKuhn's emphasisontheory choiceor
acceptance,
for in Carnap's version of epistemology,
theories
arc not chosen or
accepted
b u t
o n l y
p robab i I i f i ed .
M
Carnap's
writingsinthe1940sand
1950s
portrayhim as espousing
a logical
conception of
probability.
But by the
late 1950s
an d ear ly
1960s,
he
clearly favored
a view
that
can be t e rmed
tempered per-
sonal i sm: probability
is
rational degree
of
belief.
15
I
will
h a v e
more
to say on
this matter
insection 8, but in the
m e a n t i m e
I
will present
Carnap
as a
tempered Bayesian personalist.
A
shotgun marriage
of
Kuhn
and
Carnap
could
be
arranged
by
t ak ingCarnap
to
supply
the
probabilities,
K u h n to
supply
the
values
or
utilit ies,
and
then applying
the
rule
of
maximizing
the
expected
utilityto
render
adecisionon
theory choice.
16
But
likemost shotgun
marriages, this
one
would
be a
mistake.
For
Carnap
it
would
be a
mis take because it w o u l d
invo lve the
pretense t ha t
the
accepted
the-
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
15/29
22
John Carman
ory
is certa in even t h o u g h
one's
degree of belief in
the
theory may
be less thanone,perhaps substant ia l lyso. For K u h n it would alsobe
a mis take,
since
the
efficacy
of his values
does
n ot
depend
on the
t ru th
o f the
theories,
so
estimates
of the
probable t ruth
of the
theories
is
i r relevant
to K u h n i a n
theory choice.
Part
of the
wrang le here derives from
the
un f o r t un a t e ph ra s e
"theory choice." Scientists do choose theories, but on beha l f o f Car-
n a p , Iwould c l aim th a t they
choose
themon ly inthei n n o c uo ussense
that theychoose to devote their t ime an d energy tothem: toart icu-
la t ing
them,
to
i m p r o v i n g th e m ,
to
drawing
ou t
their consequen ces,
to
confront ing
them
withthe results ofobservationand experiment.
Choice in thissense allows for a reconciliation of
Bayes
a n d K u h n ,
since
thischoice
is i n fo rm ed bybo th Baycsian an d Kuhnian fac tors :
probabil i ty an d
th e
valu es of accuracy, consistency,scope, simplicity,
an d
f ru i t fu lncss .
Alas, this reconcil iation israther shallow. Once we are clear tha t
th e sort of choice involved in
" theory
choice" is a p ractical
one, there
is
n o t h i n g
sacred
a b o u t
the
list
of
i tems
on
K u h n ' s
list
of values.
Other
values,
such as
ge t t i ng
an
N S F
g r a n t or w i n n i n g the
No be l
Prize,
can and do
enter.F u r the r,
the
k ind
of
choice
in
quest ion a l lows
a
scientist to be bigamous, s ince he can choose to work on two or
more theories at
once,
and i t al lows him to be fickle, since he can
oscillate back and forth. The kind of choice Struc tureenvis ioned w as
m u c h more p e r m a n e n t ; indeed,
the
impression given there
is tha t
norma l
science
is not
possible
without tying Cathol ic
bonds
to a
theory,b o n d s
tha t
can o n l ybe broken b yl eav ingthe C h u r c h , i.e., by
creat ing a revo lu t ion .
Is there no way to b r idge the gap be tween Carnap an d K u h non
this
issue? T o see howba f f l ing the Bay esian f inds the not ion of theory
acceptance, con sider the case of Einstein 's gen eral theory of relativity
( G TR ) ,
a rguab ly the leading theory o f gravitation an d thus the top
candidate for acceptance. Marie, a research worker in the field fa-
miliar
with
all
of the
re levan t exper imenta l
f indings ,
does
some
in -
trospection and finds
that
her degreeof belief in GTR isp.
Case
1: p is1 or so
n e a r
1 a s makes no
odds . Here
there
is a n a t u r a l
sense in which the Bayesian can say that Marie accepts GTR. Such
cases,however,arc so
rare
as to const i tute anomalies .Ofcourse,one
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
16/29
23
C a r n a p ,
K u h n ,
an d
llic Phi losophy
of
Scientif ic Mclhodology
can cite any number of cases f rom the his tory of
science
w h e r e
scientistsseem
to besayingfor
their
pet
theories
that theysetp= 1.
Here
I
would
urge
the need
to dis t ing uish
careful ly
between
scien tists
as advocates
of
theories versus scientists
asjudges of
theories.
T he
latter
roleconcern sushere,and in
that
role
scientists
know,or should
k n o w ,
tha ton ly in v ery exceptiona l
cases
does the evidence rat ion al ly
suppor t a
full
belief in a theory. Let us move on to case 2.
Case
2: p is , say , .75. Subsequent ly Marie decides to
accept
G T R o n
the
basis
of her
probabil i ty
assignments and the
values
she attaches
to GTR and its
competitors.
W h at
could
th i s mean?
Subcase 2a .
W h e n
she
accepts GTR, Marie changes
her degree of
belief from .75 to 1.
This
i s no th ing
short
of
folly, s ince
she has
a l ready
m a d e
a considered judgment about evident ial
support
an d
no new
relev an t evidence occasioning
a
r e j u d g m e n t
has
come
in .
Suitcase 2b . W h e n
she
accepts
G T R ,
Mar ie
docs
n ot
c h a n g e
her de-
gree of
belief f rom
.75 to 1 , bu t she
acts a sz / a l l doub t were
swep t
away
in
t ha t
she
devotes eve ry w ak in g hour
to
show ing tha t var ious
p u z z l i n g
as t ronomica l
observations
can be exp la ined by the
theory ,
she
assigns her graduate s tudents research projects that presuppose
the correctness of the theory, she wri tes a textbook ongrav i ta t ion al
research
that
isdevoted
almost exclusively
to
GTR, etc.
But at this
p o i n t
w e
have come
full
circle back
to a
sense
of
theory acceptance
t h a t
is
really
a
m i s n o m e r ,
for
w h a t
isinvolved is a
practicaldecision
about the allocation of
personal
and insti tutional resources and not
a decis ion about the
cpis tcmic sta tus
of the theory.
This ra ther pedant ic d ia t r ibe on theory acceptance would b e best
forgotten were
it not for
its impl icat ions
for our
p ic ture
o f
n o r m a l
science. Aswe have seen, theory
choice
or "acceptance" can refer
cither to adopting an epislemic
attitude
or to mak ing a practical
choice. In the form er case there is n o n atu ra l Bay esian explication
of theory acceptance save in the case where the probabi l i ty of the
theory isone.
Since scientists
asjudgesoftheoriesare
almost never
in
a
position
to
ju s t i fy
such
an
acceptance,
th e
Bayesian predict ion
i s
t h a t rarely is a theo ryaccepted in the epis temicsense.S imilarly,w h e n
theory choice is a
m a t te r
ofdeciding w hat theory todevote
one'sl im e
an d energy to, the Bayesian predict ion is tha t in typical s i tuat ions
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
17/29
John
l .arnian
where members of the c o m m un i t y a s s i gn d i f f e r en t
utilities
to such
devotions,
they
will
m a k e
di f ferent choices .
Thus
from
e i ther
the
cpistcmic
or
practical-decision
perspective,
the
Bay esian
predic t ion
is
diversity.
This
predict ion
is, I
t h i n k ,borne
out by
actual
scientific
practice.
In
section
9below
Iwilla rgue that insofar
as
normal science
implies
a
shared pa rad igm,
the
pa rad igm need
not and in fact
of ten
is
not so
specific
as to i nc lude a pa r t i cu l a r
( accepted ) theory .
I
w i l l
also
hazard a
proposal
for a
m i n i m a lsense
of
"shared p a r a d i g m "
tha t
yields
a less
strai t- jacketed image
of
n o r m a l
science an d
that also
d imin i shes
wi thout ob l i t e ra t i ng
the
di f ference between normal
an d
revolut ionary
science.
By way o f closing this section an d
i n t r od u c i ng
the
nex t ,
I
will
cons ider a f inal way of reconci l ing K u h n an d C a r n a p on theory
choice. Radically n ew theories, so the story goes, carry with them
di f fe ren t l inguis t ic /conceptua l f rameworks . Thus,
even
seriously
lo
enter ta in the new theory involves the decision to
adopt,
if on ly t en -
tatively,
the new framework, and this decis ion is for Carnap a
prag-
matic on e that involves the sorts of factors emphasized i n K u h n ' s
account o f paradigm
replacement.
In
response
I would
repeat
w h a t
I have
a l ready
said insection 4: major scientific revolut ions
such
as
the
t rans i t i on f rom Newton ian to special-relativistic physics- an d
thence lo genera l rela t ivi tyneedn't be seen as forc ing a choice be-
tween i n c o m m e n s u r ab l el ing uist ic /con ceptualsystems, since
it is of ten
possible lo f i t
ihc possibilities in to
a
la rger scheme tha i m a kes
llic
theories c o m m e n s u r a b l e lo the
ex ten l
lhal
conf i rmal ion
ques t ions
can
be
posed
in t e rm sof an observat ion
base lhal
isn c u i r a l e n o u g h
for assessingihe
re la l ive conf i rmat ion
of
the
theories.
However,
ihe
rccognil ion of ihe
larger
possibil i tyset canproduce adrastic
c h a n g e
in prob abi l i ty va lues , achan ge best
described
in K u h n i a n
terms.
7 Revolutions andB elief Shifts
A mild
f orm of scient i f ic revo lut ion occurs w ithihe in l roducl ion of
a new
llieory lhal articulates possibil i t ies
thai lie
wi lh inihe b o u n d -
aries of
ihe
space of iheories to be taken seriously b ut that , because
of the fai lure of actual scientists to be logicallyom nisc ien l , had pre-
viously been unrecognized as explicil possibilities.T he more radical
form
of revolulion occursw h e n ihe spaceof possibilities itself needs
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
18/29
25
C a r n a p , K u h n , an d
ihc
Ph i losophy
o f Scicnl i f ic
Methodology
to be s ign i f ican t ly al tered to encompass
the
n ew theory . In practice
the
d is t inc t ion
between
the two forms of revolu t ion may
be b lu r r ed ,
perhaps even hopelessly so, but I w i l l begind iscussion by
p r e te n d i n g
t ha t
we can per form a
separat ion
ofcases.
Even the mi ld form of revolu t ion may induce a non-Bayes ian shi f t
in belief func t i ons . By non-Bayes i an I
m e a n
that no form of condi-
t iona l i za l ion , whether s t r ic t o rJeff rey or some natura l ex tens ion o f
these, w i l l suffice to expla in
the
ch ang e . For cond i t iona l i z i ng ( in any
recognizable sense of
the t e r m ) on the i n fo rm at ion tha t j u s t n o w a
here tofore una r l i cu l a ted
theory
T
has
been
i n t roduced
is
l i terally
nonsens ica l ,
becausesuch acon di t ion al iza t ion
presupposes
that pr ior
to this t ime there was a w el l-defined probab i l ity for th i s i n for m at ion
an d t h u sfo r T , wh ichisexactly wh at th efai lureof logical om niscience
rules
out.
W e can t ry to acknowledge the
f a i l u re
of
logical
omnisc ience b y
m e a n sof A b n e r
Sh i m on y ' s
(1970) deviceof acatch-all hy poth esis
H
f
,
which
asserts in e f fec t tha t someth ing , we know not what , beyond
th e
previously formulated theories
T\
T z,
. . . ,
T
T
is
t rue.
Now
suppose tha t a new theory T is introduced an d tha t as a result the
old degrce-of-bcl ief f u n c t i o n Pr ischanged to Pr ' . T he most conser-
vative
way the
sh if t f rom
Pr to
Pr' could take
place
is by the
process
I will
call
sha ving
o f f , n a m e l y ,Pr(T,)
= .P r ' ( T i ) for i = 1, 2, . . . ,qan d
Pr'(T) = r > 0 and
Pr'(//
e
)
= Pr(//
c
)- r.Thatis,u n d e r shav in g o f f ,
H
f
serves
as a well
for init ial
probabilities for as y et u n b o r n
theories,
an d the
ac tua l
i n t roduc t ion of new
theories resul ts only
in
d r a w i n g
u p o n
this
well w i t hou t d i s t u r b i n g th e probabi l i t ies o f
previously
for-
mu la ted theories . U n fo r tun a te l y , such conse rva ti sm even tua l ly leads
to the a s s i g n m e n t of ever smal ler
in i t ia l
probabi l i t ies to successive
waves
of n ew
theories
u n t i l a
p o i n t
isreached
where
the new
theory
has
such
a low
ini t ial probab i l i ty
as to
s tand
n ot
m u c h
of a fighting
chance .
Cer t a in ly s h a v i n g
off is a factual ly
inadequate descr ip t ion
of
w h a t
h a p p e n s
in
m an y scient i f ic revolut ions , especial ly
of the
more radical
type . Th ink o f w h a t happened fo l l owing the in t roduct ion of E i n -
stein's special theory
of
relativi ty (STR)
in
1905. Between 1905
an d
1915 little n ewempir ical evidence in favor o f S T R wa srecorded, an d
yet th e probabi l i ty of compet ing theories, such asthose of Lorentz
an d A b r a h a m , set in classical
space
an d t ime,
fell
in the estimates of
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
19/29
26
John Ea r m a n
most
o f
th e m e m b e r s
of the
E u r o p ea n p h ys ic s c o m m u n i ty ,
and the
probab i l i t y
subtracted
from
these
e lect ron theor ies
w as
t rans fe r red
to E ins te in ' sSTR.The probabi l i t iesof au xi l ia ry hypothesesmayalso
be a f fec ted , as i l lustrated by the in t roduc t ion of the general theory
of relat ivity
(GTR).
When Eins te in showed tha t
GTR
accounted
for
the exac t am ou n tof the anomalous advance ofM ercury ' s pe r ihe l ion ,
the
hypothesis of an amount of zodiacal
matter suf f ic ien t
to
affect
Merc u ry ' s per ihel ion dropped dramat ica l ly in the estimates of mos t
of the
phys ics communi ty ( see Earman
and
Glymour 1991) .
In u s i n g
the
term non-Bayesian
to describe
such noncond i t ion -
alization bel ief chan ges , w hether of the conservative shaving-off type
or
some more rad ica l fo rm,
I do not
mean
to
imp ly tha t
the
changes
are not i n f o r m e d b y
Bayesian
considerat ions . Indeed, the problem
of the t rans i t ion f rom Pr to Pr' can be t hough t of as no more and
no less than the fami l i a r Bayesian problem of ass ign ing ini tial
prob-
abili ties, on ly n ow
wi th
a new in i t ia l
si tuat ion
i nvo lv ing a newset of
possibilities
and a
n ew
in fo rm at io n bas is . B ut the
problem
we
arc
n ow f ac ing isq u i t eu n l i k ethose a lleged ly solved b yclassical pr in cip les
of
ind i f fe rence
or
modern var ian t s the reof , such
as E. T.
Jayncs 's
m a x i m u m - e n t r o p y
p r inc ip le , where
it is
assumed
w e
k n o w n o t h in g
o r
very
little ab ou t the
possibilities
in
question.
In
typical
cases the
scientific
c o m m u n i ty willpossessa vas tstoreo f relevan t exper im en ta l
an d
theore t ica l in format ion . Us ing tha t in format ion
to
i n f o r m
the
red is t r ibu t ion
of
probabil i t ies over
the
competing
theories
on the
occasion
of
thei n t r o d u c t i o n o f
th e
new theory ortheories is a
process
t ha t , in the str ict
sense
o f the t e r m ,
is\o?Slipnal:
i t cannot 'be accom-
plished
by some nea t fo rm al ru les o r ,
T o
u s e K u h n ' s
term,_bvan
a lgor i thm.
On the
otherjjand,
the
process
is far f rom
beirgjrrationa>>
since it is informecf by reasons. But the
reasons,
as
K u h n
Tias~crrf-
phasized, come in
the
f o rm
ofpersuasions
ra ther than proof. In
Bayesian
te rms ,
the
reasons
are
marsha led
in the
guise
of
ptausiblTily
a r g u m e n t s .The d e p l o y m e n tof plausibi l i tya r g u m e n tsis anart form
for which there
current ly exis ts
n o
t axonomy .
A n d
cons ider ing
the
l imitless variety of such a rgumen t s , it isun l i ke ly t h a t a n y t h i n g m o r e
t h a n
asup erf ic ia l t axon om y can be deve loped . E ins te in , the consum -
mate mas te r of th isart fo rm, appea led to analogies), symmet ry con-
s iderat ions ,
thought exper iments , heur i s t ic
principles
such
as
the
pr inc ip le
o f equ iva lence ,
etc.
All of
these
considerat ions, I am sug-
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
20/29
27
C a r n a p , K u h n , an d the
Ph i lo sophy
o f Scientific
M ethodology
gcs t ing on
be ha l f
of th e
Baycsians ,
were deployed to n udg e a ssign -
m e n t s
of
in i t ia l
probabi l i t ies in
favor
of
the
theories Eins te in was
i n t r o d u c i n g
in the
ear ly
decades
of
this cen tury . E instein 's success
in
this
regard
is no
less important than experimental evidence
in ex-
p l a i n i ng
th e recept ion of his theories.
T o s u m m a r i z e , K u h n ' s pu rp le passages do no t
seem
overb lown
w h e n
applied to revolut ions in the st rong sense d ist inguished
above.
T he persuasions that lead to the adoption of the newshape fo r
the
possibility
space can no t a m ou n t to proofs . Certa in ly for the
Bayesian
they
can no t cons is t o f indu c t iveproofs, since the very
a s signmen t
of
degrees
of
belief
presupposes
the
adoption
of
such
a
space.
After
a
revolut ion has taken place, the newand o ld theories can often be
fi t ted
i n to
a common f r ame tha t be l i e s an y vicious fo rm of i nc o m -
mensurab i l i ty (as I tried to illustrate in section 4 for N e w t o n i a n an d
relativistic theories). But this retrospective
view
tends to disguise the
shake-up in our system of beliefs occasioned by the adoption of the
n ew
shape
fo r
the possibi l i tyspace.Baycs ian i sm b r ing s
the
shake-up
to
l ight , albei t
in a way
tha t unde rcu t s
the
s t anda rd fo rm
o f
the
doctr ine .
I have no way of kn ow in g wh e the r Ca rn ap w ou ld have app roved
of my
B ayes ian read ing
o f
K u h n .
But I do
claim tha t
it is a
read ing
t ha t
fi ts
na tu ra l l y wi th Carnap 's mature v iews
on
probabil i ty
an d
induc t ion .
8 Objectivityand the Problem of
Consensus
I haveendorsed a Bayesianized version of Kuhn ' s c l a im tha t in sci-
enti f ic revolut ions persuasion ra ther than proof is the order o f the
d a y :
revolut ions
involve the in t roduct ion of new possibilities; this
in t roduc t ion causes the redistribution of probabilit ies; the redistri-
but ion isguided b y
plausibi l i ty
a r g u m e n t s ; an d such a rguments be -
long to the art of persuasion.
This
endo r semen t
is
conf ined
to the firststage of the
revolut ion,
when
the
ini t ia l
probabilit ies are established for the expanded pos-
sibility set. The Bayes ian fo lk lo re would have it that after thisfirst
stage, s o m e t h i n g
more
ak in
toproof
than persuasion operates.
The
idea
is
that
an
evidence-driven consensus emerges
as a
result
of the
Bayesian l e a rn ing model:degrees of belief change by conditional -
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
21/29
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
22/29
2 )
C a r n a p , K u l i i i , a n d th e Philosophyof
Scienti f ic
Methodo logy
a p robab i l i t y f u n c l i o n . If th i s were
corrccl, ihc liaycsian
ver s ion o f
llie
problem
o f
scicnl i f ic
objectivi ty
would
be
made
correspondingly
easier.
However,
I do not
think
that Carnapmanaged tostakeout a
d e f ens i b l e position, as can be brought out by the question of how
one recognizes wh i ch probability functionsare rational. Carnap's
a n s w e r
w as toappeal towhat
he
variously called inductive
i n t u i l i o n "
and induclivc common
sense.
The trouble, of course, isthat one
person's i n d u c t i v e common sense
is
another's i n duc t i ve non-sense.
So the appea l to i n t u i t i o n reveals
very
diffcrenl opinions as lo
whelhcr
i t is rat ional to
learn from
experience
at all
and,
if so, at
w h a t rate.
A t
th isjuncture
il
wil l
b e
he l p fu l
to
review
a
mechanism
proposed
by L e h r e i an d Wagner in
Ra t i ona l
C o n s e n s u s in Science a n d Society
(1981) for achieving agroupconsensus.Their mechanism requires
t h a t
th e
members of ihc community
change ihcir
degrees o f bel ief
in
accordance wi lh
a
wcighled-aggrcgalion
ru le .
Suppose that
a t
the
i n i t i a l
m o m e n t , person
i has a
degree
o f bel ief /;, in the theory in
question.
Each
person i is
assumed
to
assign
a
weigh t
w/
0 to
every
person
_ ; ' , w h i c h can be
taken
as an index of i ' s
opinion
as to
th e
re l iab i l i ty
of/s opinions. According toLchrcr a n d Wagner's r u l e , i
t he n improves her i n i l i a lopinion /> ? b ychanging it topi = X,w,y/>".
If there arc
still
differences of opinion, the aggregation processis
repeated
w i th
the /;,' to obtain
f u r t h e r
improved probabilities
/;?,
etc., u n t i l e v e n t u a l l y the
probabi l i t ies
for a l l
members
fall
into
l ine .
Lchrcr and
Wagner
o f f e r aconsistencya rg u m e n t for
their
aggre-
ga t ion ru le : If a
person re fuses
toaggregate,
though
hedoes
assign
a
posi t ivew e i g h t
to
other
members,he is
acting
as
though heassigned
aweightof
one
toh imse l fand aweightofzerotoevery
other
member
of
thegroup.If, in fact,he assignspositiveweightto
other
members
of th egroup, ihenheshould nolbeha veas ifheassigned zero we igh t
lo them (1981, 22).Thisargumenthas the flavor of When arc you
going
to
slop b e a t i n gyour
wife ." I d o
assign
a ,
positive weight
to the
o p i n i o n s
of
others,
but as a
liaycsian
I do
this
not by
means
of
weigh ted aggregation b u t
by c on d i t i on a l i z a t i on :
I co nd i t i o na l i zc on
i n f o r m a t i o n about the opinions o f m y
peers,
a n d I notice that th e
result
is a sh i f t in mydegreesof belief toward thedegreesof belief
of those I respect.When I was ayoungstudent,
these shif ts
brought
.m y
opinions
closely in
line
wi th
those
belonging topeople Iregarded
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
23/29
3
John
lirmun
as the expert s , but as a m at ur e m em ber of the com m un i ty , I f ind
tha t
such
shifts,
whi le
still
n onn egl ig ib lc , do no t
c o n f o rm
m y
o p i n i o n s
lo
those
of
others,
a t
least
no t on mat ters
where
I now
regard myself
as an expert . A n d I res is t any further
a t t em pt
to
bend
m y
careful ly
considered opinions .
There
are tw o
reasons,
i n d e p e n d e n t
o f
Bayes ian ism,
to be
u n h a p p y
with Lehrer an d
Wagner 's proposal
an d ones like i t . The first is
t h a t
it is
descriptively
false, as shown by the very
e x amp l e
they
use
to
motivate the i r proposal. In the 1970s Robert Dicke claimed that
optical
measurement s
of the
solar disk revealed
an
oblateness large
enou g h to accoun t for 3" to 5" of arc in Mercury ' s centenary per i -
helion ad v ance an d thus to throw in to
doubt
Eins te in ' s explanat ion
of
th e
a d v a n c e . W h e n
o th e r
as trophysicis ts disagreed with Dicke ' s
conclusions the
d i f ferences were
n o t
smoothed over
b y
p r od u c ing
a
consensual probabi l i ty
by
means
of a
w eighted aggreg ation process.
The d i s a g reem en t
remains unreso lved
to
this day.
T he
weight
of
op in ion
docs
seem
lo
be
go ing
aga ins t
D icke's in lcrp rc ta t ion , but
this
part ial ag r e e me n t
is in
fact
due not toag gregation but to the acq ui-
sition
of
addi t ional ev idence .
Of course,Lchrcr an d W a g n e r are
perfectly
aw ar e of these facts,
an d
the
descr ip t ive
inadequac ies
of the i r proposal do no t concern
t hem,s ince they
take
them selves to be off er in g a n orm ativeproposal.
B ut
even
in
ihcsc t e r ms
the
proposal should
be
f au l t ed .
It is
f u n d a -
m en ta l
lo science thai opin ions be evidence-dr iven . Differences of
opin ion
need
not
constilutc
an
emb arrassmen t that
needs
lo
be
quashed ,
for ihesc
d i f fe rences
can
serve
as a
spur
lo
fu r the r
iheo-
relical
an d experimental research, and the new in fo rmat ion p ro -
duced may d r ive a gen u in e scientific consensus . The
al ternat ive,
an
a t l cmpl lo m a n u f a c lu r e aconsensus by a w eighted-aggrega tion pro-
cedure, smacks
of the
"mob psychology"
of
which
K u h n w as
criticized.
This
last point generalizes. Bayesianism
an d
other approaches
to
scienlific in fe rence
aswell
suggest that unless
there is
some evidence-
dr iven
process that
operates
on the level of ind iv idualscientists to
produceagroupconsensus ,ihe
consensus willa m o u n l
lo
someth ing
tha t ,
if not mob psychology, is nevertheless a social art ifact not de-
serving ei ther of the labels "rational" or "scienlific."
Thus,
con l rary
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
24/29
C a r n a p ,
K u l i n ,
a n d
the
Phi losophy of Scientif ic Methodology
to K u h n ' s idea,
th egroupc a n n o t
decideit
can no t r a t ion a l ly dec ide
to
agree if
the i n d i v i d u a l s
disagree.
I do not see how
th is con clus ion
can be escaped
u n l e s ssome
y et tobe a r t icula ted
collcct ivisl m e th o d -
ology is show n lobe viable .
9 A
Partial Resolution
to the Problem ofConsensus
Part of the
an s w e r
to the Bayes ian
version
of the
problem
of
consen-
sus isthat q ui te of ten itdoes n o t exist an d does n o t
need
toexist for
normal sc ient i f ic
research
to
take place.
S t ruc t u r e
warned
of the
d a n -
ger of
t ak ing t ex tbook science as our image of how real
science
actual ly operates,
and in
par t icu lar ,
it
showed
how
textbook science
t e n d s lo m a k e scienti fic r evo l u t i on s
i nv i s ib le
b y p a i n t i n g an ov e r l y
rosy p ic tureof asm ooth ly ac cu m ula t in g s tockp ileof scientific k n o w l -
edge. B u t
I t h i n k t h a t St ruc ture fai led to emphas ize how textbook
science also disguises
the
d ivers i ty of opin ions and approaches that
f lour i sh in n o n r c v o l u t i o n a r y science.
If I had the space, 1w o u l d
o f f e r
as a case s t u d y the
d e v e l op m e n t
of relativisticg r av i t a t ion a lresearch over the last
seven ty-f ive-years .
19
Textbooks
in th is
area
have t ended to
be
books on E in s t e i n ' s G TR ,
t h u s fos ter ing
th e i l lusion t ha t
G T R h a sachieved
the
s tatus of
para -
d igm
h e g e m o n y .
In addi t ion , ear ly textbooks n o t on ly d ow np lay ed
the exis tence o f r ival theories b u t disguised ser ious difficul t ies w i t h
tw o
of
the
p r i nc ip a le x p e r i m e n t al tests ofG T R , the red
shif t
an d
the
b e n d i n g
o f
l ig h t . N o r m a l
scientif ic
research
in
this
field
c o n t in u e d
in
the face of
bo th
a cha l lenge lo the th i rd exper imenta l
leg
o f G T R
d e r i v i n g from D icke' ssolar-oblatenessm easurem en ts and a lso an ever
g r o w i n g n u m b e r o f
r ival theories
o f
gravi tat ion.
This and s im i l a r
e x a m p l e ssuggest
t h a t n o r m a lscience
is possiblew h e n thec o m m u n i ty
ofexperts
share
a
parad igm
in the
weak sense
of
ag r eemen t
on the
ex p l ana t o r y d oma in
of the field, on the
c i rcumscr ip t ion
of the
space
of
possible theories
to be
considered
as
ser ious candidates
for
cov-
ering
the
ex p l ana t o r y d oma in ,
on
exemplars
of
explanatory success,
and on key aux i l i a ryhypotheses. (I am
t empted
to say
tha t this
is the
m i n i m a l
sense o f parad igm needed to
u n d e r w r i t e
normal sc ience ,
b u t h is tor ians o f sc ience probably have counterexamples w a i t i n g in
thewings . )
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
25/29
John Kaim an
One could a rgue t ha t n o t h a v i n g a pa rad igm in th e stricter sense
of
a shared
theory
of
gravi ta t ion has
lowered
the puzzle-solving
efficiency
of normal science. One can recal lThornca n d Will 's (1971)
l a m e n t
thai , faced wi th
a zoo of
a l tern a t iv e theor ies
of
g r a v i t a t i o n ,
astrophysicists where
hamstrung in thei r model -bui ld ing activity.
While I
t h i n k t ha t this
is a fa ir
observa t ion ,
I
also
t h i n k
tha t there
is
more
toprogressinn o rm a l
science
tha n puzzle solving. In particular,
I would
emphas ize
the conceptua l advances derived f rom
the
explo-
ration of the
space
of
possible theories,
a
po in t tha t br ings
me to the
second part
o f my
par t ia l answer
to the
problem
of
consensus.
Again , if I had the space, I wo uld argue tha t in sofar as a consensus
is
established, it is often due to a
process
akin to the much
maligned
idea of e l imina t ivc i n d u c t i o n . T h is process isoften accomp an ied b y
a proliferation of
theories,
not as an exercise in
Feyerabendian
an-
a rchy
or
D a d a i s m ,
but as a
m e a n s
of
p ro b i n g
the
possibilities
and as
a p r e l i m i n a r yto deve lop ing a classification scheme tha t ma kes sys-
tematic
e l imina t ion
a t ractable
exercise.
20
T he
e l i m i n a t i o n
is no t
of
the s i m p l c m i n d c d Sher lock Holmes var ie ty ,
for it
involves
B ayesian
e lements , especially
in the
assessment
of the
auxi l i a ry a s sumpt ions
needed to b r i n g a b o u t a conf ron ta t ion o f theory an d e x p e r i m e n t .
Thus
the Baycsianizcd versionof the
problem
ofconsensus
remains .
An d a t the p r e sen t
t i m e
I do n ot see any
reso lut ion tha tdoes
n ot
f a l l
back
on s o m e t h i n g l ike the def in i t iona l so lu t ion , which I casually
dismissed in th e preceding sec t ion . Such a fallback un de rm i n es sci-
en t i f i c
objectivi ty
in a way tha t
wo u l d
n o t
have pleased
C a r n a p
a n d ,
I
p r e s u m e , docs n o t please
K u h n
c i t he r .
10
Conclusion
I
was a
d is tan t
s t u d e n t of C a rn a p and a c lose s tudent of K u h n .B u t
the two seemed to me so d i f f er en t in s tyle and concerns that I placed
them in d i f f e ren t pa r t s of the phi losophical f i rmament . O n l y n ow
have
I begun to apprec ia te how mi sgu ided m y p l acemen t was and
how m u c h phi losophyof science can be
enr iched
b ycons ider ing how
the ideas of these tw o gian ts in terac t . I have
presented
one way to
stage the in terac t ion .There are su rely bet ter ones. Iurge more able
hands
to
take
up the task.
21
-
8/11/2019 Earman - Carnap, Kuhn, And the Philosophy of Scientific Methodology
26/29
33
C a r n a p , K u l i n , an d
th e
Phi losophy
of Scientific
Methodology
Notes
Sections6-9
of
th is
paper arc