You are on page 1of 11

ANDREAS KAMLAH

AN I M P R O V E D DEFINITION OF
'THEORETICAL IN A G I V E N T H E O R Y '

J. D. Sneed offers a definition of theoretical terms of a theory, which


might roughly be characterized as terms introduced by a new theory not
yet belonging to the previously available language. This definition is
defective in many ways.
Sneed's criterion says roughly (see Sneed 1971, p. 31-33): A function n
is theoretical in T i f f from the description of measuring n in a given
application of T follows always that T has at least one application. One
may raise immediately the following objection: The criterion is too
strong, since at least in the case of theoretical mechanics we cannot even
derive for any function that the theory is applicable at all. For instance the
description of a scale (for determination of masses) does not involve a
dynamical theory but only statical considerations. It is equally well
coinpatible with Aristotelian dynamics (with the force law m 9v = f) as
with Newtonian dynamics.
Sneed of course has an answer to this standard objection: How do we
know that we are measuring just the mechanical mass and not something
different from it, if not every mass measuring method presupposes that
Newtonian Mechanics has at least one application (Sneed argues in a
similar way on page 117 of his book, Sneed 1971)? Sneed does not show
however, that his requirement is not too strong, though he is indeed
pointing out correctly, that any mass measuring method has to satisfy a
condition, from which its consistency with methods directly drawn from
mechanics can be derived. Thus there might still be another method of
characterizing a concept as theoretical in a theory or intimately related to
a theory, against which the above-mentioned objection cannot be raised
and which can be shown to be more adequate, as for instance the one
presented below.

Erkenntnis 10 (1976) 349-359. All Rights Reserved


Copyright (~ 1976 by D. Reidel Publishing Company, Dordrecht-Holland
350 ANDREAS KAMLAH

A proponent of Sneed's criterion could argue here, that we have


not paraphrased it correctly. Sneed uses the expression " . . . t h e
d e s c r i p t i o n . . . p r e s u p p o s e s . . . " and not " . . . from the descrip-
t i o n . . , f o l l o w s . . . " as we do. One could with Stegmfiller interpret
the word 'to presuppose' in a weaker sense 1. We do not want to deny
that a skillful interpretation - the simplest would be to replace 'presup-
poses' by 'agrees with' - might remove the apparent difficulties of the
criterion. For sake of logical analysis however one might prefer to direct
one's objections against a univocal, more rigid and possibly also more
vulnerable version, even if one is not sure that it was intended by the
author.
We do not want here to give a detailed interpretation of Sneed's book
or of his opinions, but rather propose an alternative improved explication
of 'theoretical', which is clear enough to need no additional interpretation
and may well serve as a basis of his systematic exposition of the 'logical
structure of mathematical physics'. Thus Sneed's in some respect reason-
able account of physics is put onto a firmer basis. The main idea of our
approach is, that the conceptual system of physics is built up together with
an operational hierarchy of theories starting from low level theories,
which are quite near to immediate experience, and ascending to high level
theories quite remote from every-day-life physics. Physical concepts are
obtained successively in this course along with new theories. Mechanics
for instance enables us to define mass if our previous language contains
only kinematical concepts. From electrodynamics we may obtain electri-
cal charge on the next level of the operational hierarchy. Thus mechanics
is sufficient for defining the measurement of mass, not necessary in any
(and not even in a weaker) sense.
For our explication we start with a physical language containing
nonlogical predicates and functions ~bz.A description of measurement of a
function ~bi is a statement Mn(~bl, ~b2,..., ~bn), where the nonlogical
constants appear in the arguments of Mn. We may interpret M~ as a
relation between sets which at least partially determines the extension of
~b~, if the extensions of ~bl. . . . . ~br are known. Thus if ~bl,. . . . . ~br are
already measured or known by direct observation, the relation Mn tells us
at least something about the extension of ~br+l. . . . . ~bn, or in favourable
cases determines it completely. We write the statement made by the
theory in the same way as T~(~bx,..., ~bn).
AN IMPROVED DEFINITION 351

We now call a function theoretical in Tn with respect to the functions


61 . . . . . 6~ iff every description of measurement M,(61, 6 2 , . . . , 6,),
(n/> r), follows logically from Tn and from a statement M~'(61. . . . ,6~), a
description of measurement of 61 . . . . ,6~ (r<-n). The conjunction of
these is also one of the M2(61,..., 6r). Therefore there is an Mr', such that
allM, follow from Tn and from M~iffr~+l . . . . . 6 , is theoretical relative
to 61 . . . . . 6~. Due to the fact that the M~" are elements of a boolean
lattice, the order in which the quantifiers appear here is irrelevant.
Written in a semiformal way with the existential quantifier as the first
symbol:
6~ is theoretical in T, with respect to 6 1 , . . . , 6,-1 iff

V A (T.(61 ..... 6.) ^ M;(61 ..... 6~)


M'M~

(1) ~- M,(61 . . . . ,6n)).

Thus we can say that the theory Tn defines the functions 6r+a . . . . . 6 , if
61 .... , 6r are given2. Essential for our definition is the reference to the
vocabulary of the language already available on the next lower level of
the physical hierarchy. This hierarchy may be constructed in different
ways, one may for example interchange the levels of mechanics and
geometrical optics, construct either the mechanical functions before the
optical or vice versa. Therefore reference to a level in the hierarchy is only
possible by mentioning the vocabulary of this level explicitely. Thus we
do not claim that there is a meaning of 'theoretical term in T,' in an
absolute sense as Sneed does, but merely relative to a given language.

A t this point the following question m a y arise: W h a t about those functions, which are
neither theoretical nor belong to the set of functions {4~1. . . . . 4~r}, relative to which
{~br+l. . . . . 4~,} are theoretical? Is not the set {4,1. . . . . ~b,} already the vocabulary of the
relative observational language? T h e answer is: Formally we m a y take any set of functions
and ask, which other functions are theoretical in relation to them. T h e r e will be then a
9 possibly n o n e m p t y class of nontheoretical functions not contained in the reference set
{q~l. . . . . ~br}, (which is of course trivially nontheoretical). But if we are looking for a suitable
relative observational language for our theory, as for instance the kinematical language is
for mechanics, all functions not contained in the reference set have to be theoretical,
otherwise this set cannot be called a relative observational language. The reason is, that we
need for those third-class functions measuring operations which are neither given by the
theory itself nor belong to the already known operations of the set {~bi. . . . . ~,}, which is the
352 ANDREAS KAMLAH

basis of our physical language, in the situation described above where we try to build up an
operational hierarchy. Thus our criterion tells us for which basic sets ~bx. . . . . ~b, we can
extend the physical language by the use of a theory alone without the help of other auxiliary
theories.
We may also ask, if our criterion satisfies Sneed's above mentioned requirement, that
theoretical functions have to be the functions defined by the theory. I think it is in full
agreement with all relevant intuitive comments, which Sneed gives to his explication of
'theoretical' (see Sneed 1973, p. 117).

The advantage of this explication of theoreticity will become dear, if we


will have shown, that it satisfies a very important criterion of adequacy, a
condition expressing the Ramsey eliminability of theoretical terms. We
shall formulate this criterion for the special case, where only one term is
eliminated. The generalization to the case of several functions
~br-1. . . . . ~b, is straightforward. By Ramsey elimination of ~b, we mean
here the replacement of T,(cbl . . . . . ~b,) by Vx T,(~bl . . . . ,4~,-1, x). This
is justified, if we can derive from T,(~bl,..., ~b,) no more and no less
empirical facts than from Vx 7", ((~,..., d~,,_~,x), or if these two formula-
tions are empirically equivalent. We try to make this idea more explicit
now:
It is not difficult to define the concept of empirical equivalence for
contingent statements, which are not contained in the corpus of physical
laws. Two such statements are empirically equivalent, if each of them in
conjunction with all known physical laws leads to the same empirical
conclusions. For the characterization of empirical equivalence of
nomological statements we have to modify this concept if we do not want
to accept the consequence that all nomological statements are empirically
equivalent to tautologies. We do this in the following way: Two nomo-
logical statements are empirically equivalent, iff from the conjunction of
each of them with the remaining physical laws the same empirical
conclusions have to be drawn.
It is difficult however to characterize the set of the remaining physical
laws in a unique and language independent way. Instead of trying to solve
this difficulty we prefer to talk of the empirical equivalence of two
nomological concepts relative to a certain statement T~',which represents
these remaining laws. We shall now apply these considerations to the
concept of Ramsey eliminability of a term, which we now conceive as the
empirical equivalence of a theory T, with its Ramsey sentence
V~ T,(4~1 . . . . . ,~,,_~, x).
AN IMPROVED DEFINITION 353

The function r is Ramsey eliminable relative to the rest of


physics T', iff the confunction T. A T[ contains the complete
physical knowledge about the functions r 1. . . . . r and for all
expressions A.-1(r . . . . . r the statement
(2)
V Tn(r Cn--1,x)AT~'(r . . . . . r ~- A n - l ( r . . . . . 4,-1)
x
is equivalent to
T~(dfl,,..., ~bn) A Tnr(~bl, 99 9 ~b.) b-- An-,(O1 . . . . , ~b,-,) 9

This criterion may be shown by purely logical transformations to be


equivalent to a second condition:

For the already mentioned statement T" holds

V T.(r . . . . . r x) A V ~L'(r 99 9 r y) F-
x y
(3)
V (Tn(r .... , r X) A T n / ( r ..... Cn--1, X))
x

which says, that the existential quantifier may be put before T. and T"
separately.
We may now prove easily that our definition of theoreticity is sufficient
for satisfying conditions (2) or the equivalent condition (3). For this
purpose we have to look a bit more closely at the theories of measurement
M. and M;. These are relations between extensions of physical functions
(/~1, 9 9 ~bn. The knowledge of some of these (not necessarily r . . . . . &r)
determines at least in some cases partially the extensions of the other
functions. At the same time the M , and Mr' are of nomological character;
they have to be accepted laws of physics. Since this is all we demand, for
the theories of measurement, every physical statement of nomological
character may serve as such a theory. If we remember this, we see by a
short derivation that our condition (2) follows from the definition (1). As
in the definition of Ramsey eliminability we restrict ourselves in this
derivation to the case of r = n - 1. Applying the logical formula

A (a (x) -> b (x)) ^ V a (x) -->V b (x)


x x x
354 ANDREAS KAMLAH

we obtain from definition (1):

V A (V T.(~a . . . . ,4.-a, x) ^ M.'-a(~l . . . . . ~._1)


M~_I Mn x
(4)
VM,,(4,~,..., ~.-1, x)).
x

In the next step we have to identify m t n _ l ( t ~ l , . . - , 0 n - l ) with


Vx T'(~bl . . . . , ~b,_l, x). On what grounds is such an identification possi-
ble? From M ' - I in conjunction with Tn we can derive all measuring
theories M,. Since there is nothing special about a measuring theory, the
set of these theories is nothing but the whole available physical know-
ledge about ~ba. . . . , ~b,. Therefore M ' - a must be a statement containing
the rest of the physical knowledge about ~b1. . . . ,4~,, which is not already
contained T,(~bl,...,~b,), and thus satisfies all conditions for T'.
Since it does not depend on ~bn, the existential quantifier Vx does
not change anything. Identification of M'-a(4~l,...,~b,-1) with
V y Tn(dgl . . . . , c~,,_1, y) and substitution of V~ M.(4h . . . . . (~n-1, X) by
V~ (T. (4h . . . . . ~b._a, x) ^ Tn(~ba. . . . , tP.-1, x)) yields directly condition
(3), which is equivalent to condition (2), which is now valid relative to a
r~(q~a,... , ~b,) for which Vr T'(tka, 9 9 9 ~b,_~, y) = M'_l(~ba . . . . . th,-1).
It seems to be clear, that the Ramsey eliminability of theoretical terms
has to be shown by a proof similar to this one. Sneed does not show for his
criterion that such a proof works. In this respect our criterion is superior
to Sneed's and may be used to save his approach to physics from the
breakdown of his criterion as already indicated above.
But there remains still one difficulty, which has to be resolved in this
paper if our explication is to be of any relevance for Sneed's account of
physics. Sneed's formalism looks very different from the one used here
and we have to show that this difference is only apparent. Given a physical
theory in the reconstruction of Sneed, which differs from his fully
developed scheme only by not accounting for the difference of theoretical
from nontheoretical terms, we have to be able to arrive at Sneed's final
form for the theory in question by the aid of our criterion. Thus we
become involved with Sneed's formalism explicitly, without being able to
give a brief outline of his reconstruction scheme of physical theories in a
few pages. (For such a review see Stegmiiller, 1975.) Therefore the
AN IMPROVED DEFINITION 355

following paragraphs presuppose acquaintance with Sneed's book. We


shall simply refer to his formulas as he develops them in his text.
Though for Sneed theories are concepts and not statements, he does
not claim that the physicist states nothing about the world. There has
indeed to be something capable of being true or false in physics, if physics
is an empirical science. For this purpose Sneed introduces the theory
statement I e N~ (Sneed, 1971, p. 182, D37), which says, that I, the range
of application of the theory falls under N~, the concept of its mathematical
structure. Sneed makes fairly clear what is meant by N~. It is more difficult
however to understand the nature of I, the range of intended applications,
the function of which is twofold: On the one hand I refers to the
extensions of the physical functions in the real world, which describe the
physical systems, to which the theory is applied. On the other hand I
includes only the range of those physical systems, for which the physicist
expects the theory to be a fairly good approximation of a possibly
unknown exact theory, which would be valid for an unlimited range. We
want to separate here the two roles of I by splitting up this set into two
components:
(5) z=I~nI~.
The second component IR selects the range of those systems, for which
the physicist would use the theory as a good approximation. The first
component ls connects n-tuples of abstract mathematical functions (in
Sneed's terminology the elements of No, see p. 165, D25) with sets of
physical functions of our world, which may be characterized by their
description of measurement. In our account these functions are given by
&l . . . . . ~br. To be more exact: an xcNo is an element of ls iffx is an
n-tuple consisting of a set of particles Nx, a time interval Tx, the Cartesian
product of which may be called the domain Dx, and moreover of the
extensions of ~b1. . . . , ~br,which are functions on Dx. (In the case of wave
theories, which have a different ontology, Dx is a region of space-time, in
which the physical process described by x takes place.)
We now introduce a relation x = y, which will be seen to be useful to
express Is in our formalism. If x and y are elements of No, that is,
descriptions of possible physical systems used by the theory, if Dx and Dy
are their domains, then x ___y is defined by
(6) XEy := Dx __GDyA{X, y}~ C=,=.
356 ANDREAS KAMLAH

Here {x, y} e C== is Sneed'sway of expressing that the functions in x and


y have to be mutually consistent in the region, where Dx and Dy overlap
(see p. 88-89, and p. 170-171; on p. 88 C== is written as ( = , = )). Thus
x r--y means intuitively, that the physical process described by x is a
mereological part of that described by y, or a part in space-time. We can
now write x ~ Is taking Dw to be the maximal domain, which contains all
others, the whole world:
(7) x~I~:=xr-'(Dw, dp~. . . . . dpr).
The ~bx,..., ~br are logical constants denoting the real extensions of
physical functions defined by their descriptions of measurement. Thus
x ~Is in our interpretation says, that an element of No, which is a
set-theoretical entity, is connected with reality in a certain way, and it is
because of this connection that the theory statement can be empirically
true or false. This statement now takes the form:

(8) {xlxz(Dw, d~l. . . . ,qbr)}nIR~N~.


We see immediately, that this is a statement of the form Tr(4~a. . . . , ~b,)
for which we have tried to discuss the problem of theoretical terms.
If we, however, look more closely at Sneed's formalism, we see, that N~
is a set of subsets of No, I a subset of No and that the elements of No
contain in their n-tuples only nontheoretical terms. It would be pointless
to look for theoretical terms in a theory of the form of Equation (8), since
this expression is already the result of the separation between theoretical
and nontheoretical terms and contains only terms of the latter type as
logical constants. Therefore we have to replace I and N~ by sets con-
structed on the basis of n-tuples, which also contain the theoretical
functions. Sneed calls the set of these untruncated n-tuples Mo (p. 165,
D25). The elements of Mo are related to those of No by the restriction
function r, which truncates them:
(9) No=R(Mo) with R(Y) := {x]V(y e Y ^ x = r(y))}.
y

We replace now I by I,h, a subset of Mo. These two ranges of application


are related by I = R (Lh). The theory concept N, is replaced by M~, a set of
subsets of Mo, which is related to N, by:

(10) N~ = { X I V ( Y ~ M ~ ^ X = R ( Y ) ) }
Y
AN IMPROVED DEFINITION 357

The 'theoretical' theory statement is then Lh ~ M~, and if we split up Ith in


the same way as we have done with I into Isth and 1R,h, Ith = Is~h C~1Rth,
ls= R(1sth), IR= R(1Rth), we have:

(11) T.(r . . . . . r r ..... r


This expression differs from expression (8) by referring also to the
theoretical functions Cr+, . . . . , r (n t> r.) To the T thus defined we may
apply the criterion of our paper, in order to find out which of the functions
~bx. . . . . Cn are theoretical in relation to our previous language; we then
obtain the Ramsey sentence of Tn:

(12) T,(,/,1,..., ,/,,)~ V . . . V Tn (,/,1. . . . , ,/,, x , + , , . . . , x~).


Xl, rl + 1 Xn

If we want to use our criterion for theoretical terms as justification for


replacing the 'theoretical' theory statement by Sneed's theory statement,
which accounts for the difference of theoretical and nontheoretical terms,
we have to show that the Ramsey sentence of Equation (12) is identical
with the expression of Equation (8). Otherwise it would be pointless for
Sneed's formalism, that we have proved the Ramsey eliminability of all
theoretical terms in the statement formalism. This is indeed possible; but
we do not want to prolong this paper by another proof. For the reader it
may also be more important to understand the content of the theorem
then to follow the technical details of a proof, which is only a chain of
symbolic transformations. The theorem is the following:

IfforlRth holds: (a) Iath ={x]V (y =r(x) ^y ~-IR)}


Y

(b) IR=R(Imh); (C) lath ~ C=,=, and/f Tn(r . . . . . &n)

is defined by Equation (11), then the statements of expression


(8) and of both sides of the logical equivalence (12) are
equivalent.

The proof, which is not given here in detail, proceeds as follows: From
expressions (10) and (8) we obtain:

(13) V (I~nIR=R(Y)^ Y~M~).


Y
358 ANDREAS KAMLAH

By making use of the conditions (a), (b) and (c) for I we may transform
this expression into the equivalent:

(14) V(Is=R(Y) AY~IRthEM~AYEC=,=).


Y

The term Y ~ C== is superfluous in this expression, since it is already


implied by the rest. What remains to show is, that for any expression
~ ( . . . ) the following equivalence holds:

V (Is = R (Y) Alse C=,= A ~ ( Y ) ) ~--~


Y
(15)
V . ' " V ~'~({XIx r'-(Vx, ~ 1 . . . . . ~ m , X m + l , " 9 9 , Xn)})
Xm +l Xn

By this last step, where we have to make use of some properties of C==,
the proof is completed.
Thus it can be shown, that our results and our definition of theoreticity
are indeed important for the logical analysis of physical theories, even in
the approach of Sneed. The statement view and the non-statement view
are not so much different as it may appear at first sight. Different views are
here not different positions but merely ways of looking at the same thing
from different sides and thus observing different aspects of it.

ACKNOWLEDGEMENTS

The author wishes to thank Dr Wolfgang Balzer and Dr George Berger


for valuable discussions and for controlling formulas and English style.
Fachbereich 6 (Mathematik/ Philosophie )
der Universit?it Osnabriick

REFERENCES

Sneed, J. D.: The Logical Structure of Mathematical Physics, Dordrecht (Holland), 1971.
Stegmiiller, W.: Theorie und Erfahrung, 2. Halbband (Studienausg. Teilbde, D u. E);
Berlin-Heidelberg-New York, 1973.
Stegmiiller, W.: 'Structures and Dynamics of Theories', Erkenntnis, 9 (1975), 75-100.
AN IMPROVED DEFINITION 359

NOTES

1 Stegmiiller points out, that the difficulty to understand how we 'presuppose' Kant's
synthetic a priori principles is very similar to the difficulty, which arises in this context
(Stegmiiller, 1973, p. 62).
2 It would be convenient to improve our definition by excluding terms from being
theoretical, which are defined by a noncreative axiom. We could do this, following Majer
('Uber die adfiquate Methode der Einfiihrung von theoretischen Begriffen'-to be pub-
lished), by an additional condition. Since the modified definition defines only a narrower
concept, we would still be able to prove the Ramsey eliminability for it.

You might also like