Logique &

Logique & Analyse 164 (1998),

Analyse 164 (1998), 327-342
327–342

VAN FRAASSEN'S
VAN FRAASSEN’S CONSTRUCTIVE
CONSTRUCTIVE EMPIRICISM,
EMPIRICISM, SYMMETRY
SYMMETRY
REQUIREMENTS AND SCIENTIFIC REALISM
A ND SCIENTIFIC

MICHEL GHINS
MICHEL GHINS

Abstract. With
Abstract. Within
in the framework of a semantic view of science, such
as the
as the constructive
constructive empiricism advocated by
empiricism advocated by van
van Fraassen,
Fraassen, it
it is
is shown
shown
that the
that the empiricist
empiricist position
position naturally leads to
naturally leads to symmetry
symmetry principles
principles inin
the construction
the construction of
of models
models in in physics.
physics. AAss aa consequence
consequence of those
of those
symmetry constraints, itit is
symmetry constraints, is argued
argued that
that aa selective
selective form
form of
of realism
realism
for theoretical
for theoretical objects
objects is
is more
more plausible
plausible than
than antirealism.
antirealism. Several
Several
examples drawn
examples drawn from
from spacetime
spacetime physics,
physics, static
static and
and mechanics
mechanics are are
discussed, particularly
discussed, particularly the
the transition
transition from
from Cartesian
Cartesian to
to Newtonian
Newtonian
mechanics.

The semantic
The semantic or or model-theoretic
model-theoretic approach
approach of of theories
theories hashas long
long become
become the the
dominant view
dominant view inin philosophy
philosophy ooff science.
science. I Itt opposes
opposes the the logical
logical positivist
positivist
or syntactic
or syntactic conception
conception of of scientific
scientific theories.
theories. Fo Forr the
the latter,
latter, aa theory
theory is is es-
es-
sentially
sentially aa set
set ooff theoretical
theoretical postulates
postulates oorr axioms
axioms whose
whose theoretical,
theoretical, non- non-
observational, terms
observational, terms areare interpreted
interpreted by by means
means ooff analytic
analytic correspondence
correspondence
rules which
rules which connect
connect eacheach theoretical
theoretical term
term toto aa set
set ooff observational
observational terms,terms,
i.e. terms
i.e. terms which
which directly
directly refer
refer to
to observable
observable entities,
entities, properties
properties or or processes.
processes.
The objections
The objections to to this
this neo-positivist
neo-positivist view have been
view have been abundantly
abundantly put put forward
forward
in the literature (see for example Suppe (1989), van Fraassen
in the literature (see for example Suppe (1989), van Fraassen (1980)) and (1980)) and itit
would
would be pointless to rehearse them. After a brief reminder of the central fea-
be pointless to rehearse them. After a brief reminder of the central fea-
tures
tures ofof the
the model-theoretic
model-theoretic approach,
approach, II will give some
will give some precisions
precisions on on the
the no-no-
tion of
tion of model
model which
which will be used,
will be used, mainly
mainly inspired
inspired from
from vanvan Fraassen's
Fraassen’s con- con-
structive empiricism which,
structive empiricism which, in in accordance
accordance with actual practice
with actual practice in in physics,
physics,
gives a prominent role to symmetries in model
gives a prominent role to symmetries in model construction. I willconstruction. I will then
then
show that this particular brand of the model-theoretic construal
show that this particular brand o f the model-theoretic construal of science of science
naturally leads
naturally leads toto restrictions,
restrictions, inin terms
terms ooff symmetry
symmetry requirements,
requirements, on on thethe
construction of models. As a consequence of these constraints,
construction of models. A s a consequence of these constraints, it is argued it is argued
that scientific
that scientific realism
realism is is more
more plausible
plausible than
than antirealism,
antirealism, in in particular
particular in the
in the
domain of classical spacetime
domain of classical spacetime physics. physics.
II wish
wish to
to thank
thank Harvey
Harvey Brown,
Brown, Sílvio
SIlvio Chibeni,
Chibeni, Dennis
Dennis Dieks,
Dieks, Steven
Steven French,
French, Jan
Jan Govaerts,
Govaerts,
Décio Krause, Bas van Fraassen for their useful comments on a first draft of this paper.
Décio Krause, Bas van Fraassen for their useful comments on a first draft of this paper.
328 M I C H E MICHEL
L GHINS
GHI NS

1. The
The notion
notion of model

The fundamental
The fundamental sense,sense, that
that played
played aa role
role of of paramount
paramount importance
importance in in the
the
hands of the heirs of the logical positivists (Putnam and
hands of the heirs of the logical positivists (Putnam and Quine, among oth- Quine, among oth-
ers) is
ers) is the
the well-known
well-known logicallogical sense:
sense: aa model
model is is what
what satisfies
satisfies oror makes
makes truetrue
aa set
set of propositions. This sense, although obviously still
of propositions. This sense, although obviously still accepted and used accepted and used
by the
by the proponents
proponents of of the
the semantic
semantic view,
view, slid
slid to to the
the background
background among among thosethose
who advocate a model-theoretic approach of science
who advocate a model-theoretic approach o f science because it is closely because it is closely
linked with
linked with the
the neo-positivist
neo-positivist conception
conception which,which, among
among other other drawbacks,
drawbacks,
laid too much emphasis on language and axiomatization.
laid too much emphasis on language and axiomatization. According to According to the
the
defenders of the model-theoretic approach, a theory is not primarily a set of
defenders of the model-theoretic approach, a theory is not primarily a set of
propositions or
propositions or laws,
laws, butbut only
only derivatively,
derivatively, to to the
the extent
extent that
that such
such proposi-
proposi-
tions describe
tions describe thethe structure
structure ofof aa model,
model, or,or, rather,
rather, aa class
class ofof models: models
models: models
are prior. Models moved to the forefront, they “occupy
are prior. Models moved to the forefront, they "occupy center stage" as center stage” as van
van
Fraassen puts it. A model then should not be introduced
Fraassen puts it. A model then should not be introduced by starting from a by starting from a
set of axioms but should be presented “directly”, independently
set of axioms but should be presented "directly", independently of the laws of the laws
and axioms which
and axioms which they they may
may make
make true
true and
and which
which could even be
could even be unknown.
unknown.
Thus, the
Thus, the model-theoretic
model-theoretic or or semantic
semantic conception
conception of of theories
theories is is not
not only ad-
only ad-
verse
verse toto the
the positivistic credo that
positivistic credo that issues
issues in in the
the philosophy
philosophy of of science
science areare inin
fact issues in the philosophy of the language of science
fact issues in the philosophy of the language of science but also to the con- but also to the con-
tention that
tention that philosophy
philosophy of of science
science is is first
first ofof all concerned with
all concerned with the
the logical,
logical,
syntactic and even semantic, properties of axiomatized
syntactic and even semantic, properties of axiomatized theories. theories.
Of course, aa model
Of course, model can can only
only be be conveyed
conveyed in in aa language
language (this
(this is,
is, as
as van
van
Fraassen remarks,
Fraassen remarks, aa "trivial
“trivial point").
point”). Wh Whata t isis aa model
model then?
then? LLet me quote
e t me quote
here aa quite
here quite revealing
revealing footnote
footnote from
from van
van Fraassen's
Fraassen’s Laws Laws andand Symmetry:
Symmetry:
“In my
"In my terminology
terminology here
here the
the models
models are
are mathematical
mathematical struc-
struc-
tures called models of a given theory only by virtue of
tures called models of a given theory only by virtue o f be- be-
longing to the class defined to be the models of that theory”.
longing to the class defined to be the models of that theory".
(1989, p.
(1989, p. 366)
366)

2. The relation
2. The relation of
of models
models to phenomena
to phenomena

A physical
A physical theory
theory cannot
cannot bebe only
only aa set
set ooff purely
purely mathematical
mathematical structures
structures
but must entertain an —at least possible— relationship to phenomena.
but must entertain an —at least possible— relationship to phenomena. This This
relationship is characterized by van Fraassen as “embedding”:
relationship is characterized by van Fraassen as "embedding": observable observable
phenomena are
phenomena are embedded
embedded —or —or embeddable—
embeddable— in in models. Phenomena which
models. Phenomena which
are
are isomorphic
isomorphic to parts or
to parts or portions
portions of
of the
the models
models areare said
said to
to be
be embedded
embedded in in
them. In other words, phenomena are shown, by abstracting
them. I n other words, phenomena are shown, by abstracting some of their some of their
aspects
aspects or contents, to
or contents, to possess
possess aa structure
structure which
which is is identical
identical to
to the
the structure
structure
of some parts of the models. The parts of a model
of some parts o f the models. Th e parts o f a model M which are M which are isomor-
isomor-
phic to
phic to possible
possible phenomena
phenomena are are called
called "empirical
“empirical substructures"
substructures” E. These
E. These
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 2 9 329

substructures,
substructures, being being subsets
subsets ooff mathematical
mathematical models,
models, must
must obviously
obviously also
also
be mathematical. Given that isomorphism means complete
be mathematical. Given that isomorphism means complete identity of struc- identity of struc-
ture, itit follows
ture, follows that,
that, for
for an
an empirically adequate theory,
empirically adequate theory, observable
observable phenom-
phenom-
ena have the same mathematical structure as the empirical
ena have the same mathematical structure as the empirical substructure: substructure: only
only
their contents or “matter” may be
their contents or "matter" may be different. different.
“Empirical adequacy is
"Empirical adequacy is truth
truth with
with respect
respect to
to the
the observable
observable phenomena"
phenomena”
(van
(van Fraassen 1989, 192-193). Tru th here means isomorphism of
Fraassen 1989, 192–193). Truth here means isomorphism of an em-
an em-
pirical substructure with the actual, real, structure of phenomena (which are
pirical substructure with the actual, real, structure of phenomena (which are
not sense
not sense data).
data). van
van Fraassen
Fraassen supplements
supplements his his model-theoretic
model-theoretic view view with a
with a
correspondentist, and even representational, conception of
correspondentist, and even representational, conception of truth. Models are truth. Models are
not only
not only copies
copies ooff possible
possible worlds,
worlds, but
but possible
possible worlds
worlds themselves
themselves in in the
the
sense that they possess and exhibit a formal structure which
sense that they possess and exhibit a formal structure which can in principle can in principle
be shared,
be shared, if the model
if the model isis correct,
correct, by
by the
the real
real world.
world. Empirical
Empirical substructures
substructures
of theories which save the phenomena correspond (and
of theories which save the phenomena correspond (and here correspondencehere correspondence
means complete
means complete identity
identity ofof mathematical
mathematical structure) to real,
structure) to observable struc-
real, observable struc-
tures of the phenomena.
tures of the phenomena.

"“(...) the empirical structures in the world are the parts which
(are at
are at once
once actual
actual andand observable"
observable” (1989, (1989, p.
p. 228)
228)
.
It follows
It follows
. that phenomena
that phenomena embedded embedded in in empirical
empirical substructures
substructures (this
(this is
is
aa limiting
limiting. case
case of
of embedding
embedding when
when a
a model
model narrows
narrows down
down to
to an
an empirical
empirical
substructure)
substructure)
) have mathematical
have mathematical structures
structures (also
(also called
called by by van
van Fraassen
Fraassen ap-
ap-
pearances
pearances t
(1980,
(1980, p.
p. 64)).
64)). The
The operation
operation of
of embedding
embedding reveals
reveals or
or exhibits
exhibits the
the
real structures
real structures in in phenomena.
phenomena. Th This
is view bears
view bears aa striking
striking resemblance with
resemblance with
h
some
some of of Wittgenstein’s
Wittgenstein's ideas ideas inin the
the Tractatus,
Tractatus, where
where aa true
true proposition
proposition hashas in
in
common e a logical form with the reality it depicts (a proposition functions as
common a logical form with the reality it depicts (a proposition functions as
“scaleemodel”).
aa "scale model"). It It seems
seems aa direct
direct consequence
consequence of of van
van Fraassen’s
Fraassen's construc-
construc-
tive m
tive empiricism
empiricism that an
that an empirical
empirical substructure
substructure has has aa mathematical
mathematical form form inin
common
common with the
p with the phenomena
phenomena itit represents.
represents. ThisThis appears
appears particularly
particularly clearly
clearly
from one
from onei ofof van
van Fraassen's
Fraassen’s mainmain examples,
examples, the the seven-point
seven-point geometry,
geometry, which
which
occurs both
occurs both
r in The
in The Scientific
Scientific Image
Image (p. (p. 42)
42) and
and Laws
Laws and
and Symmetry
Symmetry (p. 219).
(p. 219).
In this
In this well-known
iwell-known example,
example, aa specific
specific structure
structure ofof lines
lines and
and points
points made
made
of, say,
of, say,cropes
ropes andand nails
nails (the
(the "matter"
“matter” ooff the the phenomena),
phenomena), is is embedded
embedded in in
Euclidean
Euclidean a space.
space.
Thus,
Thus,l inin physics
physics at at least,
least, a a theory
theory is is aa set
set of
of mathematical
mathematical structures
structures which
which
contain empirical
contain empirical substructures
substructures and and which
which exhibit
exhibit symmetry
symmetry and and invariance
invariance
s
properties.
properties.t
There r is quite a large array of mathematical structures as candidates for
There is quite a large array o f mathematical structures as candidates for
scientific u theories.
theories. TheThe mainmain difficulty
scientific difficulty in in physics, as in
physics, as in other
other mathematized
mathematized
sciences,
sciences, lies in constructing models that have some chance of
c lies in constructing models that have some chance of being
being empir-
empir-
ically adequate,
t that is, which have a possible connection
ically adequate, that is, which have a possible connection with phenomena. with phenomena.
u
r
e
s
i
n
330 M I C H E MICHEL
L GHINS
GHI NS

The mere
The mere consideration
consideration of
of mathematical models themselves
mathematical models themselves does
does not
not permit
permit
to draw a demarcation between the substructures which are empirical
to draw a demarcation between the substructures which are empirical and and
the ones which are non-empirical or meta-empirical. Such a distinction
the ones which are non-empirical o r meta-empirical. Such a distinction is is
intrinsic to
not intrinsic to mathematical
mathematical models
models and
and must
must then
then come from elsewhere,
namely from
namely from experience.
experience.
Having approvingly referred to
Having approvingly referred to the
the works
works of
of Przelewski,
Przelewski, Wojcicki,
Wojcicki, Dalla
Dalla
Chiara, Toraldo di Francia, Patrick Suppes, Frederick Suppe, Sneed and
Chiara, Toraldo di Francia, Patrick Suppes, Frederick Suppe, Sneed and oth-
oth-
ers, van
ers, van Fraassen
Fraassen goes
goes on
on to
to say:
say:
“Certain parts
"Certain parts of
of the
the models
models were
were to
to be
be identified [underlin-
identified [underlin-
ing is ours] as empirical substructures, and these
ing is ours] as empirical substructures, and these were were the
the
candidates for the representation [underlining is ours] of the
candidates for the representation [underlining is ours] o f the
observable phenomena
observable phenomena which
which science
science can
can confront
confront within
within our
our
experience”. (1989, p. 227)
experience". (1989, p. 227)

To
To qualify as such
qualify as such an
an empirical
empirical sub-structure
sub-structure E E must
must bebe included
included in
in at
at least
least
one model M: M:
E
E⊆ CM
AccordingM toto constructive
constructive empiricism,
empiricism, the
the aim
aim of science is
is to
to save,
save, oorr
According o f science
rather to
rather to embed,
embed, the
the observable
observable phenomena,
phenomena, that
that is
is to
to build
build up
up models
models which
which
contain substructures
contain substructures isomorphic
isomorphic to
to all possible phenomena
all possible phenomena within
within the
the realm
realm
of
of the theory.
the theory.
van Fraassen pursues:
van Fraassen pursues:
“At this
"At this point
point itit seemed
seemed that
that the
the relationship
relationship thus
thus explicated
explicated
[embedding] corresponds exactly
[embedding] corresponds exactly to to the one Reichenbach
the one Reichenbach at- at-
tempted to
tempted to identify
identify through
through this
this concept
concept of of coordinative
coordinative defi-
defi-
nitions, once
nitions, once we
we abstract
abstract from the linguistic
from the element. Thus
linguistic element. Thus inin
a space-time
a space-time the
the geodesics
geodesics are
are the
the candidates
candidates for the paths
for the paths of
of
light rays
light rays and
and particles
particles in in free
free fall. More generally,
fall. More the identi-
generally, the identi-
fied spatio-temporal
fied spatio-temporal relations
relations provide
provide candidates
candidates for
for the
the rela-
rela-
tional structures
tional structures constituted
constituted byby actual
actual genidentity
genidentity and
and signal
signal
connections” (1989,
connections" (1989, pp.pp. 227-228).
227–228).

Thus,
Thus, iiff science
science aims
aims at
at constructing
constructing mathematical
mathematical models
models that save the
that save the
phenomena and if empirical adequacy is ascertained on the basis of observa-
phenomena and if empirical adequacy is ascertained on the basis of observa-
tion, we
tion, we must
must be be able:
able:

1. To identify
1. To identify in
in the
the models
models the
the parts
parts which
which count
count as
as empirical
empirical substruc-
substruc-
tures or possible representations of phenomena.
tures or possible representations of phenomena.
2.
2. TTo empirically ascertain
o empirically ascertain the
the isomorphism,
isomorphism, that
that is
is actual
actual identity
identity ooff
mathematical structure, of the structure of phenomena with the empirical
mathematical structure, o f the structure of phenomena with the empirical
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 3 1 331

substructures.
substructures.

The
The twotwo questions
questions are are tightly
tightly related.
related. AAn empiricist can
n empiricist can only provide aa
only provide
solution to the first problem on the basis of a
solution to the first problem on the basis o f a solution to the second. Fosolution to the second. Forr
example, geodesics (the metrically extremal paths)
example, geodesics (the metrically extremal paths) are empirical substruc- are empirical substruc-
tures because
tures because they they possibly
possibly represent
represent the the paths
paths ofof light rays. This
light rays. This possibility
possibility
is guaranteed by the ascertainability, in appropriate
is guaranteed by the ascertainability, in appropriate circumstances, that circumstances, that ob-
ob-
served
served light rays actually follow (or don't) geodesics. I f they do, it shows
light rays actually follow (or don’t) geodesics. If they do, it shows
that the
that the paths
paths of of light rays are
light rays are isomorphic
isomorphic to to geodesics
geodesics of of space-time.
space-time. This This
cannot be achieved, within the model-theoretic approach, by fiat or only by
cannot be achieved, within the model-theoretic approach, by fiat or only by aa
definition à la Reichenbach, but by observations.
definition a la Reichenbach, but by observations. We must be able to "see", We must be able to “see”,
given the
given the meanings
meanings of of the
the words
words "light
“light ray"ray” and
and "geodesic"
“geodesic” (according
(according to a
to a
specified
specified metric)metric) that that the
the path
path ofof aa light
light rayray is
is aa geodesic.
geodesic. For For someone
someone who who
is willing
is willing to to defend
defend the the model-theoretic
model-theoretic view view ofof science
science likelike van
van Fraassen,
Fraassen,
phenomena, to the extent that they are embedded
phenomena, to the extent that they are embedded in models, must actually in models, must actually
possess a mathematical structure. This seems
possess a mathematical structure. Th is seems to be clear if we follow van to be clear if we follow van
Fraassen when he compares embedding with
Fraassen when he compares embedding with the relationship that Reichen- the relationship that Reichen-
bach sought
bach sought to to identify
identify by by means
means ooff his his coordinative
coordinative definitions “once we
definitions "once we
abstract
abstract from from the the linguistic element”.
linguistic element".
What are
What are thethe "actual"
“actual” and and "observable"
“observable” structures
structures in in this
this case?
case? EEvenve n
though a mathematical structure is not perhaps
though a mathematical structure is not perhaps per se observable, we must per se observable, we must
be brought
be brought to to see
see oorr observe
observe that
that somesome phenomena
phenomena have have aa mathematical
mathematical
structure.
structure. I f so, as a consequence of the correspondence view of truth,
If so, as a consequence of the correspondence view of truth, those
those
mathematical structures
mathematical structures are are real.
real. IIff it is true
it is true that
that light rays follow
light rays geodesics,
follow geodesics,
itit is
is also
also true
true that
that light
light rays
rays possess
possess the the mathematical
mathematical structure
structure typical
typical ooff
geodesics. Since the structure of a geodesic is defined by means of metrical
geodesics. Since the structure of a geodesic is defined by means of metrical
relations we
relations we cannot
cannot escape
escape from
from the
the conclusion
conclusion that that light
light rays
rays manifest
manifest aa real
real
metrical structure.
metrical structure.

3. Objections
3. Objections

Let’s pursue
Let's pursue geometrical
geometrical examples
examples aa little
little further.
further. Phenomena
Phenomena in in this
this case
case
are given by
are given by the
the well-known
well-known empirical behaviour of
empirical behaviour of rods and clocks
rods and deter-
clocks deter-
mined on
mined on the
the basis
basis ooff spacetime
spacetime coincidences
coincidences ooff (quasi-)punctual
(quasi-)punctual events.
events.
Once the meanings
Once the meanings ooff geometrical
geometrical terms
terms (like
(like congruence,
congruence, oorr equality
equality of
of
lengths and
lengths and angles)
angles) have
have been
been specified,
specified, experience
experience teaches
teaches usus that
that those
those
phenomena are
phenomena are embeddable
embeddable (locally)
(locally) in
in aa manifold
manifold endowed
endowed withwith aa metric.
metric.
In the case of Minkowski spacetime, the model is an infinite
In the case o f Minkowski spacetime, the model is an infinite manifold manifold ooff
points endowed
points endowed with
with aa four-dimensional metric of
four-dimensional metric of negative
negative signature
signature andand on
on
which the Lorentz-Poincaré
which the Lorentz-Poincaré symmetry
symmetry group
group acts.
acts. Wh
What are the
a t are the empirical
empirical
substructures? They simply
substructures? They simply areare the
the intervals
intervals andand angles
angles which
which cancan be
be mea-
mea-
sured by rods
sured by rods and
and clocks
clocks (or
(or equivalent
equivalent measuring devices). O
measuring devices). Off course
course not
not
332 M I C H E MICHEL
L GHINS
GHI NS

all possibilities are

all possibilities are actualized;
actualized; not not all intervals and
all intervals and angles,
angles, etc. etc. areare marked
marked
by actually observed bodies or processes. But
by actually observed bodies or processes. But at the level of empirical sub-at the level of empirical sub-
structures, there is no reason not to accept the existence
structures, there is no reason not to accept the existence of the metric since of the metric since
itit is
is exactly
exactly whatwhat is is shared
shared by by empirical
empirical substructures
substructures and and phenomena.
phenomena.
At least two objections may be raised
At least two objections may be raised here. The first would behere. The first would be toto retort,
retort,
in a Reichenbachian guise, that the geometry, i.e.
in a Reichenbachian guise, that the geometry, i.e. the metric, is relative to the metric, is relative to
aa previous
previous coordinative
coordinative definition of congruence and that
definition of congruence and that we are confronted we are confronted
with
with aa (paradigmatic
(paradigmatic and and abundantly
abundantly discussed)discussed) case case of of underdetermination
underdetermination
of theories by phenomena. Here,
of theories by phenomena. Here, however, we would have however, we would have to to deal
deal with
with dif-
dif-
ferent empirical substructures, each corresponding
ferent empirical substructures, each corresponding to a different definition to a different definition
of congruence
of congruence (Reichenbach
(Reichenbach 1957) 1957) and and not
not the
the embedding
embedding of of the same em-
the same em-
pirical substructure in different models. Surely, the truth (or falsehood) of
pirical substructure in different models. Surely, the truth (or falsehood) of aa
statement depends on
statement depends on both
both the
the meanings
meanings of of words
words and and thethe observable
observable phe- phe-
nomena. IIff the
nomena. the empirical
empirical meaning
meaning of of the
the word
word “congruence”
"congruence" is is changed,
changed, the the
structure
structure of the description of phenomena will also have to be modified to
of the description of phenomena will also have to be modified to
preserve
preserve truth. truth. ThisThis merely
merely showsshows that that phenomena
phenomena do do not
not impose
impose aa uniqueunique
way
way of of describing
describing and and embedding
embedding them. them. Granting
Granting this doesn’t commit
this doesn't commit one one
to antirealism.
to antirealism. Moreover,
Moreover, this this kind
kind of of discussion
discussion is is certainly
certainly too too language-
language-
oriented
oriented for the model-theoretical
for the model-theoretical approach. approach.
Second,
Second, and more seriously I think, itit can
and more seriously I think, can bebe argued
argued that that we we don't
don’t ob-ob-
serve
serve the metric but only the behaviour of metrical devices or even only the
the metric but only the behaviour of metrical devices or even only the
(quasi-)coincidence
(quasi-)coincidence of of spacetime
spacetime events. events. SuchSuch aa radical
radical empiricist
empiricist position,
position,
which
which is is not
not vanvan Fraassen's,
Fraassen’s, implies
implies an an extremely
extremely narrow narrow construal
construal of of what
what
experience or
experience or observation
observation is. is. ThThe main rejoinder
e main rejoinder to to such
such aa radical
radical empiri-
empiri-
cism is
cism is to
to resort
resort to to our
our daily
daily conviviality
conviviality with with phenomena
phenomena and and the the common
common
practice of scientists. To admit that we can observe
practice of scientists. To admit that we can observe —or at least empirically —or at least empirically
ascertain—
ascertain— the the congruence
congruence of of intervals
intervals does does not
not amount
amount to to the
the desertion
desertion of of
the empiricist camp but the adoption of a more tolerant posture toward what
the empiricist camp but the adoption of a more tolerant posture toward what
can count
can count as as experience
experience or or observation.
observation. Now, Now, congruence
congruence is is mathematically
mathematically
expressed in
expressed in the
the model
model by by means
means of of the
the metric tensor. IIff phenomena
metric tensor. phenomena possess possess
the same
the same mathematical
mathematical structure structure as as parts
parts of the models,
of the models, phenomena
phenomena show, show, or or
can be made to be showing after an appropriate
can be made to be showing after an appropriate training, that their structure training, that their structure
is Minkowskian
is Minkowskian and and therefore
therefore that that the the metric
metric field
field exists.
exists. TTo oppose this
o oppose this
would
would be be tantamount
tantamount to to embracing
embracing antirealism
antirealism at at the
the level
level ofof phenomena.
phenomena.
What about
What about the the unobserved
unobserved parts parts of of spacetime? Admittedly, what
spacetime? Admittedly, what precedes
precedes
doesn’t commit
doesn't commit us us to
to the
the existence
existence of of spacetime
spacetime points,
points, nor nor to to the
the continu-
continu-
ity ooff the
ity the manifold,
manifold, but but only
only to to thethe existence
existence ooff aa metric
metric at at the
the spacetime
spacetime
locations where
locations where actual
actual phenomena
phenomena occur. occur. Th The
e useuse ooff aa continuum
continuum can can bebe
considered as an expedient device to perform
considered as an expedient device to perform the embedding. On the basis the embedding. On the basis
of the
of the previous
previous considerations,
considerations, we we must must remain
remain agnostic
agnostic with with respect
respect to to the
the
existence of
existence of aa continuous
continuous or or discontinuous
discontinuous spacetime spacetime point-manifold.
point-manifold. BBut ut
should
should we we remain agnostic with
remain agnostic with respect
respect to the existence
to the existence of of aa metric
metric in in unin-
unin-
vestigated spacetime locations?
vestigated spacetime locations?
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 3 3 333

Any arbitrary large

Any arbitrary large —but
—but ooff finite
finite size—
size— region,
region, devoid
devoid of of matter
matter and and
energy, o f spacetime can in principle be explored by rods and clocks, oorr
energy, of spacetime can in principle be explored by rods and clocks,
traversed by
traversed by free
free particles
particles and
and light
light rays.
rays. Emp
Empty regions ooff spacetime
ty regions spacetime are are
well-defined
well-defined within the model (Minkowski spacetime) without any need for
within the model (Minkowski spacetime) without any need for
reification
reification of of those
those regions
regions or or their
their metrical
metrical structure.
structure. ButBut then
then what
what is is the
the
difference between
difference between "occupied"
“occupied” and and "non-occupied"
“non-occupied” regions regions of of spacetime
spacetime as as
far as their structure is concerned? None, in my opinion.
far as their structure is concerned? None, in my opinion. Th e y all share They all share
the same
the same Minkowskian
Minkowskian structure.
structure. The The main
main point
point II want
want to to make
make is is that
that itit
is impossible in the present case to distinguish the empirical
is impossible in the present case to distinguish the empirical substructures substructures
from the
from the non-empirical
non-empirical ones. ones. AAny finite portion
n y finite portion ooff Minkowski spacetime
Minkowski spacetime
counts as an empirical substructure, i.e. is a candidate for
counts as an empirical substructure, i.e. is a candidate for the representation the representation
of phenomena.
of phenomena. The The only
only —and
—and unavoidable—
unavoidable— underdetermination
underdetermination present present
is the
is the underdetermination
underdetermination ooff the the observab/e
observable relative
relative toto the
the observed.
observed. Th The e
acceptance of a theory-model always involves a certain
acceptance o f a theory-model always involves a certain amount of extrap- amount of extrap-
olation, even
olation, even at at the
the empirical
empirical level.
level. AA given given rod,
rod, say,
say, cancan inin principle
principle be be
observed at any spacetime location and thereby render
observed at any spacetime location and thereby render observable its local observable its local
structure. (In fact,
structure. (In fact, wewe know
know that
that in in our
our physical
physical world
world Minkowski
Minkowski geometry
geometry
is approximately
is approximately correctcorrect only
only in some regions
in some regions andand for phenomena which
for phenomena which can can
be accounted
be accounted for for without
without resorting
resorting to to aa non-vanishing
non-vanishing curvature
curvature (see(see Ghins
Ghins
and Budden (2001))).
and Budden (2001))). Recall
Recall that
that vanvan Fraassen
Fraassen has has no
no qualms
qualms in in accepting
accepting
the existence of observable objects like dinosaurs.
the existence of observable objects like dinosaurs. I f so, there seems If so, there seems to to be
be
no reason for him to deny the existence of Minkowskian structure, even in
no reason for him to deny the existence of Minkowskian structure, even in
empty regions,
empty regions, since
since they
they count
count as as the
the observable
observable formform of of empirical
empirical sub- sub-
structures.
structures.

4. The
4. The role
role of
of symmetries
symmetries

Given its close

Given its close lin
linkk with visual perception,
with visual perception, geometry
geometry has has always
always played
played aa
privileged role
privileged role in
in science,
science, particularly
particularly inin physics,
physics, since
since its
its origins.
origins. TThe
he
reach of
reach of geometry
geometry has has considerably
considerably beenbeen enlarged
enlarged by by Félix
Félix Klein’s
Klein's Erlanger
Erlanger
program (1872)
program (1872) according
according to to which
which geometry
geometry isis the
the study
study o off symmetry
symmetry or or
invariance groups acting on space (which can be extended
invariance groups acting on space (which can be extended to spacetime). to spacetime).
Symmetry
Symmetry and and invariance
invariance clearly
clearly are
are key
key concepts
concepts forfor model
model construction
construction inin
modern physics.
modern physics.
The first step
The first step toward
toward embedding
embedding phenomena
phenomena is abstraction (van
is abstraction (van Fraassen
Fraassen
1989, p. 234).
1989, p. 234). ThThe same phenomena
e same phenomena can can be
be embedded
embedded in in several,
several, differ-
differ-
ent ways, depending on which aspects of the phenomena
ent ways, depending on which aspects of the phenomena are considered as are considered as
belonging to the contents and which features are taken to be formal
belonging to the contents and which features are taken to be formal or struc- or struc-
tural. Once
tural. Once thethe formal
formal aspects
aspects have
have been
been abstracted,
abstracted, we we have
have obtained
obtained an
an
empirical structure
empirical structure and
and we
we can
can proceed
proceed to to the
the embedding
embedding of of phenomena
phenomena in in
wider
wider model-theoretic
model-theoretic structures.
structures. If
I f we
we only
only have
have set-theoretical
set-theoretical inclusion
inclusion inin
334 M I C H E MICHEL
L GHINS
GHI NS

mind, an
mind, an empirical
empirical structure
structure E can be
E can be included
included in
in aa very large class
very large class of
of mod-
mod-
els M: just put in the model ingredients —additional objects, for
els M: just put in the model ingredients —additional objects, for example— example—
which
which may have nothing
may have nothing to to do
do with
with the
the phenomena
phenomena at at stake.
stake. But
But such
such moves
moves
are
are devoid of interest and van Fraassen does put constraints on the construc-
devoid of interest and van Fraassen does put constraints on the construc-
tion ooff models
tion models which
which rest
rest on
on pragmatic
pragmatic considerations,
considerations, among
among which
which the
the
predictive capacity of the theory is prominent. Symmetries play
predictive capacity of the theory is prominent. Symmetries play a decisive a decisive
role in
role in characterizing
characterizing thethe empirical
empirical content
content and
and the
the predictive
predictive power
power of
of aa
theory.
theory.
“A reflection
"A reflection onon the
the possible
possible forms
forms oorr structures
structures definable
definable
from joint experimental outcomes yields constraints
from joint experimental outcomes yields constraints on on the
the
general form of the models of the theories ‘from below’;
general form of the models of the theories 'from below'; that that
class ooff models
class models can
can then
then be
be narrowed
narrowed down
down byby the
the imposi-
imposi-
tion of postulated general laws, symmetry constraints,
tion o f postulated general laws, symmetry constraints, and and
the like, ‘from above’.” (1989, p. 228).
the like, 'from above'." (1989, p. 228).

As
As aa constraint
constraint toto model
model construction,
construction, II propose
propose the
the following
following principle,
principle,
which
which II call the Principle
call the Principle of
of invariance
invariance of
of the
the empirical
empirical substructures
substructures (PIES):
(PIES):

PIES: The
PIES: The symmetry
symmetry group
group of
of the embedding model
the embedding model must
must leave
leave its
its empirical
empirical
substructures invariant.
substructures invariant.

This
This principle
principle imposes
imposes anan integration
integration of of the
the empirical
empirical substructures
substructures within
within
the models
the models in in aa stronger
stronger sense
sense than
than mere
mere set-theoretical
set-theoretical inclusion
inclusion since
since itit
requires aa relation
requires relation between
between the
the symmetries
symmetries ooff the
the empirical
empirical substructures
substructures
and the symmetries
and the symmetries of of the model as
the model as aa whole.
whole. MoMore
re precisely, PIES implies
precisely, PIES implies
that the
that the symmetry
symmetry groupgroup E E ooff the
the empirical
empirical substructures,
substructures, which
which may
may bebe
called the
called the empirical
empirical symmetry
symmetry group,
group, cancan be
be larger
larger or
or equal, but not
equal, but not smaller
smaller
than the
than the symmetry
symmetry groupgroup S of the
S of the embedding
embedding models:
models:
E⊇
E S
D S

IIff we allowed the

we allowed the group
group S to
to be
be larger than E,
larger than then S would
E, then would contain
contain transfor-
transfor-
mations which
mations which do
do not
not leave
leave the
the form
form of
of the
the empirical substructures invariant.
empirical substructures invariant.
In such
In such aa situation,
situation, for
for any
any empirical
empirical substructure
substructure E there exists
E there exists aa transfor-
transfor-
mation s E
mation that:
∈ S such that:
sE #= EE
And sE
And may fail
sE may fail to
to accommodate
accommodate the the phenomena.
phenomena. Suppose
Suppose E is embedded
E is embedded
in aa particular
in particular model
model M M. . ThThen the transformed
e n the transformed model
model sM may not
s M may not be
be
empirically adequate
empirically adequate since
since it may not
it may not contain
contain E as an
E as an empirical substructure.
empirical substructure.
To attempt to
To attempt to clarify
clarify the
the issue
issue without
without becoming
becoming too
too technical, let’s take
technical, let's take
aa simple
simple example,
example, drawn
drawn again
again from
from geometry. Suppose the
geometry. Suppose the empirical
empirical sub-
sub-
structures are invariant
structures are invariant under
under the
the group
group ooff rigid motions (translations
rigid motions (translations and
and
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 3 5 335

rotations) which
rotations) which is is the
the symmetry
symmetry group group of Euclidean geometry.
of Euclidean geometry. IIff we we tol-
tol-
erate models the structure of which is invariant
erate models the structure o f which is invariant under the whole group of under the whole group of
linear transformations, some of these models won’t won't be able to accommodate
some
some observed
observed Euclidean
Euclidean structures:
structures: the the length
length of of an
an interval,
interval, forfor instance,
instance, is is
not generally conserved under similarities, i.e.
not generally conserved under similarities, i.e. transformations which map transformations which map aa
triangle into
triangle into any
any triangle
triangle with
with thethe same
same angles.
angles. What What would count as
would count as empir-
empir-
ically true for a model would no longer be so for
ically true for a model would no longer be so for another isomorphic model. another isomorphic model.
Thus, the
Thus, the symmetry
symmetry group group of of the
the entire
entire model
model can can be be smaller
smaller or or equal
equal but but
not larger than its empirical
not larger than its empirical subgroup. subgroup.
Moreover,
Moreover, in in accordance
accordance with with the the spirit
spirit o off an
an empiricist
empiricist approach,
approach, it it is
is
natural to demand that the symmetry group
natural to demand that the symmetry group of the model cannot be smallerof the model cannot be smaller
than the
than the empirical
empirical subgroup.
subgroup. Consider
Consider classical
classical mechanics.
mechanics. Newton Newton arguedargued
in favour
in favour of of the
the existence
existence of of anan absolute
absolute space space whichwhich contains
contains an an invariant
invariant
centre (occupied
centre (occupied by by the
the centre
centre of of gravity
gravity of of the
the solar
solar system).
system). The The invariance
invariance
group of Newton’s mechanics is a smaller group
group of Newton's mechanics is a smaller group than the invariance group than the invariance group
of Newtonian classical mechanics of point masses, since it includes only ro-
of Newtonian classical mechanics of point masses, since it includes only ro-
tations around
tations around aa privileged
privileged point,
point, the the “centre”
"centre" of of the
the world,
world, and and not
not rotations
rotations
around
around any any arbitrary
arbitrary point. point. IItt is
is well-known
well-known that that no no empirical
empirical way way exists
exists toto
find the location
find the location of of this putative centre,
this putative centre, which
which is is better
better keptkept out
out of
of the
the realm
realm
of physics.
of physics. (See (See for for example
example Weyl Weyl 1963 1963 p. p. 74,
74, Friedman
Friedman 1983 1983 pp.pp. 153-154).
153–154).
In general, the downsizing of the invariance
In general, the downsizing o f the invariance group o f the model will group of the model will lead
lead
to the addition of invariant objects (properties and
to the addition of invariant objects (properties and relations) and statements relations) and statements
since
since asas the
the size
size of of the
the group
group decreases
decreases the the number
number of of its
its invariants
invariants increases.
increases.
If we
If we admit
admit thatthat itit is
is aa desirable
desirable feature
feature of of theory construction not
theory construction not toto intro-
intro-
duce new
duce new objects
objects the the relation
relation ofof which
which to to experience
experience may may be be problematic,
problematic, it it
is safer not to allow models the symmetries
is safer not to allow models the symmetries of which form a smaller groupof which form a smaller group
than the
than the symmetry
symmetry group group of of its
its empirical
empirical substructures.
substructures. This This proposal
proposal is is not
not
to be taken as an absolute constraint but rather as
to be taken as an absolute constraint but rather as a guide to model construc- a guide to model construc-
tion, in
tion, in conformity
conformity with with anan empiricist
empiricist position.
position. Together Together with with PIES,
PIES, this this
requirement implies
requirement implies that that the
the empirical
empirical symmetry
symmetry group group be be identical
identical to to the
the
symmetry
symmetry group group of of the
the entire
entire model.
model. II thus thus propose
propose the the following
following guiding
guiding
principle, which
principle, which may may be be called
called the the Principle
Principle ooff symmetrysymmetry for for model
model con- con-
struction (PS):
struction (PS):

PS: The
PS: The symmetry
symmetry group
group of
of the model must
the model must be
be identical
identical to its empirical
to its empirical sym-
sym-
metry group
metry group

As an illustration
As an illustration ofof this principle, which
this principle, which II take
take as
as a
a local
local constraint,
constraint, let
let
us
us discuss a simple example drawn from statics. Suppose that we
discuss a simple example drawn from statics. Suppose that we set
set up,
up, in
in
aa laboratory,
laboratory, thethe following
following device
device (see
(see Figure
Figure 1)
1) with
with pulleys,
pulleys, strings
strings and
and
bodies and
bodies and that
that we
we observe
observe that
that the
the bodies
bodies don't
don’t move
move relatively
relatively to
to each
each
other and
other and that
that the
the pulls
pulls cancel
cancel each
each other
other out.
out.
336 M I C H E MICHEL
L GHINS
GHI NS

4
4 5 3

Figure
Figure 11

It can readily
It can readily be
be shown
shown that
that these
these observable
observable phenomena
phenomena are
are embeddable
embeddable
in a vector space where bodies exert pulls which can be isomorphically rep-
in a vector space where bodies exert pulls which can be isomorphically rep-
resented by
resented by vectors
vectors (see
(see Figure
Figure 2).
2).

F5

F4
F3

F5

Figure 2
Figure 2

True,
True, wewe directly
directly observe
observe pulleys,
pulleys, strings,
strings, pulls,
pulls, bodies,
bodies, etc.
etc. and
and not
not vec-
vec-
tors. I It
tors. t isis by
by abstraction
abstraction that
that we
we isolate
isolate inin the
the phenomena
phenomena or or the
the data
data aa
structure. Here
mathematical structure. Here the
the data
data are modelled in terms of vectors and
vectorial sums. Th
vectorial sums. This data structure
is data structure is
is isomorphic
isomorphic to to an
an empirical substruc-
empirical substruc-
ture which
ture which is is aa part
part of
of aa vectorial
vectorial space.
space. If,
If, as
as van
van Fraassen
Fraassen contends,
contends, atat the
the
level of
level of phenomena
phenomena “empirical adequacy is
"empirical adequacy tantamount to
is tantamount to truth”,
truth", observation
observation
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 3 7 337

provides evidence
provides evidence in in favor
favor of the existence
of the existence of of forces, i.e. vectors
forces, i.e. vectors which
which cancan
be measured by means of an apparatus like the one in
be measured by means of an apparatus like the one in Figure 1. Vectors and Figure 1. Vectors and
their relations
their relations determine
determine the the empirical
empirical substructure
substructure in in this
this case
case which
which isis “at
"at
actual and
once actual and observable".
observable”. Granted,
Granted, what we have here is a clear case of
the celebrated
the celebrated thesis
thesis ofof the
the "impregnation
“impregnation of of data
data by
by theories".
theories”. Forces
Forces may
may
be
be dubbed "theoretical" entities if desired, but this does not entail that
dubbed “theoretical” entities if desired, but this does not entail that they
they
are
are unobservable.
unobservable. If I f we
we accept
accept that
that the
the device
device presented
presented above
above isis observable,
observable,
itit can
can be
be embedded
embedded in in an
an empirical
empirical and and mathematical
mathematical substructure
substructure in in which
which
pulls are represented by vectors. The state of equilibrium of the system is
pulls are represented by vectors. The state of equilibrium of the system is
mathematically expressed
mathematically expressed by by the
the statement
statement that,
that, at
at aa point, the vectorial
point, the vectorial sum
sum
of forces is
of forces is equal
equal toto zero.
zero. Therefore
Therefore forces
forces belong
belong to to the
the empirical
empirical sub-
sub-
structures of the model and are real. Far from being a
structures o f the model and are real. Fa r from being a threat to scientific threat to scientific
realism, the
realism, the impregnation
impregnation thesis
thesis supports
supports andand reinforces
reinforces it.it. Embedding
Embedding after
after
all
all means showing that phenomena possess a structure identical to a part of
means showing that phenomena possess a structure identical to a part of
aa model and that
model and that they are, in
they are, in this exact sense,
this exact sense, theorized.
theorized.

Models, together
Models, together with with their
their empirical
empirical substructures,
substructures, areare mathematical
mathematical struc-
struc-
tures which
tures which possess
possess symmetry
symmetry or or invariance
invariance properties.
properties. IIn n aa vector
vector space,
space,
linear transformations
linear transformations act act on
on vectors
vectors and and leave
leave their
their length
length invariant.
invariant. InIn this
this
sense,
sense, length is an objective property (Weyl 1963, p. 73) o f vectors. AAn
length is an objective property (Weyl 1963, p. 73) of vectors. n
Euclidean
Euclidean vectorvector space
space makes
makes truetrue some
some mathematical
mathematical propositions
propositions (the
(the ax-
ax-
ioms of
ioms of vector
vector space).
space). But But if mathematical expressions,
if mathematical expressions, which
which we we will call
will call
‘laws’, are unavoidable
'laws', are unavoidable to to characterize
characterize the the model, axiomatization isn’t
model, axiomatization isn't indis-
indis-
pensable and
pensable and is is even
even sometimes
sometimes unavailable.
unavailable. Symmetry
Symmetry and and invariance
invariance areare
relative to
relative to aa set
set ofof laws and to
laws and to aa set
set of
of mathematical objects (properties,
mathematical objects rela-
(properties, rela-
tions) which
tions) which occur
occur in in those
those laws.
laws. ItIt is
is important
important to to draw
draw aa careful
careful distinction
distinction
between the invariance properties of laws and the
between the invariance properties o f laws and the invariance properties invariance properties ooff
the mathematical
the mathematical objects, objects, such
such as as forces,
forces, fields,
fields, charges
charges andand so
so on,
on, which
which
occur in
occur in the
the laws.
laws. In In physics
physics the the symmetries
symmetries of of the
the models
models areare the
the symme-
symme-
tries of
tries of the
the fundamental
fundamental laws. laws. This
This symmetry
symmetry groupgroup then
then determines
determines thethe way
way
in which
in which the the objects
objects are are to
to be
be transformed.
transformed. In In statics,
statics, the
the law
law ofof equilibrium
equilibrium
(for n forces)
(for 71 reads:
forces) reads:
n
!
= !0
F!i =
2=1
i=1
The mathematical
The mathematical form
form of
of this
this law
law is
is invariant
invariant under
under Galilean
Galilean transforma-
transforma-
tions. Even if the components of the force vectors in equivalent reference
tions. Even if the components of the force vectors in equivalent reference
frames may vary, their magnitudes remain constant.
frames may vary, their magnitudes remain constant.
338 M I C H E MICHEL
L GHINS
GHI NS

5. Cartesian versus
5. Cartesian versus Newtonian mechanics
mechanics

In Laws
In Laws and and Symmetry,
Symmetry, van van Fraassen
Fraassen briefly
briefly considers
considers the the transition
transition from
from
Cartesian
Cartesian to to Newtonian
Newtonian mechanics.
mechanics. Cartesian
Cartesian mechanics
mechanics was was not determin-
not determin-
istic: ffrom
istic: the (supposedly
ro m the (supposedly exact) exact) values
values of of the
the initial
initial positions
positions and and veloc-
veloc-
ities of a point-mass it is not possible to univocally
ities o f a point-mass it is not possible to univocally calculate its positionscalculate its positions
and velocities at
and velocities at some other times.
some other times. IIff we
we want
want to to “functionally” accommo-
"functionally" accommo-
date the phenomena of motion (van Fraassen 1989, p. 230) and obtain bigger
date the phenomena of motion (van Fraassen 1989, p. 230) and obtain bigger
predictive power,
predictive power, we we have
have to to inject
inject additional
additional structure
structure or or "hidden
“hidden parame-
parame-
ters” in
ters" in the
the model
model (Newton
(Newton added added forces
forces to to the
the kinematic
kinematic quantities
quantities whichwhich
were
were the only ones used by Descartes), i.e. mathematical objects which, and
the only ones used by Descartes), i.e. mathematical objects which, and
this
this isis essential,
essential, behave
behave in in accordance
accordance with with specific
specific mathematical
mathematical formulae
formulae
or laws. IIn
or laws. physics, the
n physics, the requirement
requirement that that wewe want
want to to save
save moremore andand more
more
phenomena, demands, according to van Fraassen,
phenomena, demands, according to van Fraassen, a "widening o f the theo- a “widening of the theo-
retical framework",
retical framework”, that that isis adding
adding upup new
new structure:
structure: thethe forces
forces in the present
in the present
case. “This method can be described in two ways:
case. " Th is method can be described in two ways: as introducing hidden as introducing hidden
structure, or ’dually’ as embedding” (Id. p. 229).
structure, or 'dually' as embedding" (Id. p . 229). "The word "hidden" in “The word “hidden” in
“hidden parameters” does
"hidden parameters" does not not refer
refer to
to lack
lack ofof empirical access. IItt signifies
empirical access. signifies
that we
that we see
see parameters
parameters in in the
the solution
solution which
which do do not
not appear
appear in the statement
in the statement
of
of the problem." (Id., p. 230). The problem, i.e. to achieve functionality, was
the problem.” (Id., p. 230). The problem, i.e. to achieve functionality, was
solved
solved by the introduction
by the introduction of of “hidden
"hidden parameters” (forces) which
parameters" (forces) which are are indeed
indeed
accessible
accessible to to observation,
observation, as as wewe saw,
saw, in the example
in the example drawn drawn from
from statics.
statics.
Arguably,
Arguably, the invariance group of Cartesian mechanics can be
the invariance group of Cartesian mechanics can be taken
taken toto be
be
the invariance
the invariance group group ofof the
the so-called
so-called Galilean
Galilean law law ofof inertia
inertia (which
(which was was actu-
actu-
ally first formulated
ally first formulated by by Descartes)
Descartes) and and the
the Cartesian
Cartesian scalar
scalar law
law of of collisions
collisions
(for bodies):
n bodies):
(for 11
m!v =
ruff = !k
n
! n
!
>2
mi vi = m v∗i i
,i=1
i =1 i = 1i=1
Provided
Provided 7 we assume the following three hypotheses: there exists an invari-
we assume the following three hypotheses: there exists an invari-
1
ance group,
ance group,
/
spacetime is
spacetime is isotropic and homogeneous,
isotropic and homogeneous, the the infinite
infinite velocity
velocity is
is
invariant0 (for
invariant (for Descartes,
Descartes, the
the velocity
velocity ofof light was infinite),
light was infinite), itit can
can be
be shown
shown
(Berzi and
(Berzi and1 Gorini 1969) that
Gorini 1969) that the
the invariance
invariance group
group of
of the
the law
law ofof inertia
inertia alone
alone
2
is the
is the Galileo
Galileo
:
group.
group. The
The scalar
scalar law
law of
o f collisions
collisions restricts
restricts Galileo
Galileo group
group to
to
the group
the group= of of Euclidean
Euclidean rigid
rigid motions
motions (translations
(translations and
and rotations)
rotations) and
and time
time
translations.
>
translations. I In other words,
n other words, Cartesian
Cartesian mechanics
mechanics is is not
not invariant
invariant under
under
2
boosts (except
boosts (except
,
in the particular case of linear collisions in which
in the particular case o f linear collisions in which the sys- the sys-
tem is boosted
tem is boosted
7 in the direction of the initial —and final— velocities).
in the direction of the initial —and final— velocities). The The
Euclidean
Euclidean / group is also the invariance group of the structures
group is also the invariance group o f the structures considered considered
1
empirical
empirical 4
in
in Cartesian mechanics,
Cartesian mechanics, namely
namely the the purely
purely kinematical quantities.
kinematical quantities.
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 3 9 339

To the law of inertia, Newton added the force-law (the fundamental

fundamental law of
of
dynamics or Newton’s second law) and the action-reaction law:
dynamics or Newton's second law) and the action-reaction law:
F! = m!a
P
F!A
=
A= − —F!R
m
The correct,
correct,
PR deterministic, vectorial
deterministic, vectorial law law ooff collisions is deduceddeduced fro from m
those à
two laws. The suppression of the scalar Cartesian law of collisions
those two laws. Th e suppression o f the scalar Cartesian law o f collisions
does lead
does lead to to an
an enlarging
enlarging of of the
the symmetry
symmetry group group since
since the
the invariance
invariance group group
of the Cartesian law of inertia is already the Galileo
of the Cartesian law of inertia is already the Galileo group. I n fact, forces group. In fact, forces
must transform
must transform in in such
such aa wayway thatthat the
the second
second lawlaw remains
remains invariant
invariant under
under
the Galileo group, that is, forces transform like vectors
the Galileo group, that is, forces transform like vectors as already ensured as already ensured
by the
by the statics.
statics.
Functionality
Functionality is is regained
regained together
together withwith anan enlarging
enlarging of of the
the global
global symme-
symme-
try group. But this goes along with a broadening
try group. But this goes along with a broadening of the invariance group of the invariance group of of
the empirical
the empirical substructures
substructures as as well,
well, which
which is is also
also Galileo
Galileo group.
group. Actually,
Actually,
the structures
the structures ooff phenomena
phenomena observed observed in in systems
systems in in uniform
uniform relative
relative mo-mo-
tion are now all isomorphic to empirical substructures
tion are now all isomorphic to empirical substructures of Newtonian models of Newtonian models
according
according to to the
the principle
principle of of relativity
relativity of o f Galilean mechanics. AAnd
Galilean mechanics. n d forces
forces
belong, as vectors, to the empirical substructures.
belong, as vectors, to the empirical substructures. The new objects (forces)The new objects (forces)
in Newtonian
in Newtonian dynamics dynamics represent
represent and and areare isomorphic
isomorphic to to some
some observable
observable
phenomena (pulls). Forces may be added
phenomena (pulls). Forces may be added to regain functionality, butto regain functionality, but they
they
are allowed in the model only because they can be
are allowed in the model only because they can be correlated with phenom- correlated with phenom-
ena within
ena within statics,
statics, aa branch
branch of of mechanics
mechanics which which precedes
precedes (is (is logically
logically prior
prior
to) kinematics and dynamics as Kant (1970)
to) kinematics and dynamics as Kant (1970) had already emphasized. had already emphasized.
This example is
This example is illuminating
illuminating in another respect.
in another respect. ItIt shows
shows that that PSPS does
does notnot
forbid the addition of new structure and new objects
forbid the addition of new structure and new objects (forces in this case). (forces in this case).
It only
It only requires
requires that that the
the new
new objects
objects andand the
the laws
laws which
which hold hold forfor them
them do do
not lead to a downsizing of the symmetry group
not lead to a downsizing of the symmetry group o f the whole model with of the whole model with
respect to
respect to the
the empirical
empirical symmetry
symmetry group.group. Th This again is
is again is desirable
desirable from from an an
empiricist standpoint. It makes possible in principle
empiricist standpoint. I t makes possible in principle the inclusion o f the the inclusion of the
new structure
new structure within within the the empirical
empirical substructure.
substructure. Imagine
Imagine that that some
some newly
newly
introduced structure is not invariant under the empirical
introduced structure is not invariant under the empirical subgroup. I n that subgroup. In that
event, its incorporation at the level of empirical substructures
event, its incorporation at the level of empirical substructures would render would render
the theory
the theory empirically
empirically inadequate:
inadequate: transformations
transformations allowedallowed bby the theory
y the theory
would alter the empirical substructures which will
would alter the empirical substructures which will then cease to be correct then cease to be correct
representations ooff phenomena.
representations phenomena. Th The introduction ooff forces
e introduction forces not not only
only led led to
to
the modification and enlargement of the symmetry
the modification and enlargement of the symmetry group of the "Cartesian" group of the “Cartesian”
model, but
model, but to to an
an enlargement
enlargement of the empirical
of the empirical substructures
substructures as as well.
well.
340 M I C H E MICHEL
L GHINS
GHI NS

6. The
The issue of
of the underdetermination of
of theories
theories by phenomena

The symmetry
The symmetry principle
principle PS PS was
was introduced,
introduced, as as wewe saw,
saw, as as aa recommenda-
recommenda-
tion in accordance with the spirit of empiricism
tion in accordance with the spirit of empiricism and not as an absolute and not as an absolute de- de-
mand. PS is taken to be a guide for theory construction
mand. PS is taken to be a guide for theory construction rather than a strict rather than a strict
requirement. Moreover,
requirement. Moreover, PS PS does
does not
not preclude
preclude the the possibility
possibility of of embedding
embedding
the same phenomena in different, non-isomorphic,
the same phenomena in different, non-isomorphic, models, provided those models, provided those
distinct models share the same symmetries. The isomorphy of models im-
distinct models share the same symmetries. The isomorphy of models im-
plies the
plies the identity
identity of of their
their invariance group, but
invariance group, but notnot conversely.
conversely. Models Models with with
the same
the same symmetry
symmetry group group may may differ by the
differ by the laws
laws which
which hold hold in in them
them and and thethe
objects which occur in those laws. In such a case
objects which occur in those laws. I n such a case we would be confronted we would be confronted
to aa genuine
to genuine case case of of underdetermination
underdetermination of of theories
theories by by phenomena.
phenomena. Then, Then,
an honest realist
an honest realist would
would havehave to to withhold
withhold judgement
judgement about about the the truth
truth of of some
some
laws and the existence of some objects (See McMu llin 1984, p. 11, and Gh-
laws and the existence of some objects (See McMullin 1984, p. 11, and Gh-
ins 1992
ins 1992 forfor more
more on on this
this viewpoint).
viewpoint). Such Such aa situation
situation does does not not carry
carry withwith it a
it a
widespread scepticism with respect to theoretical
widespread scepticism with respect to theoretical claims about unobservableclaims about unobservable
processes in
processes in general.
general. A A painful
painful thorn
thorn in the flesh
in the flesh of the realist
of the realist would
would be be the
the
existence of
existence of aa proof
proof that
that alternative,
alternative, distinct
distinct models
models whichwhich are are empirically
empirically
equivalent for
equivalent for all
all phenomena
phenomena and and satisfy
satisfy thethe symmetry
symmetry principle
principle PS PS can can al-
al-
ways be constructed in principle. To my knowledge,
ways be constructed in principle. To my knowledge, such a proof, the burden such a proof, the burden
of which
of which rests
rests onon the
the shoulders
shoulders of of the
the antirealist,
antirealist, remains
remains to to bebe provided.
provided.
Let me stress again that the symmetry principle
Let me stress again that the symmetry principle PS only has a PS only has a local
local import.
import.
It does
It does allow
allow forfor the
the underdetermination
underdetermination of of some
some global
global properties.
properties. GlymourGlymour
(1977) and Malament
(1977) and Malament (1977) (1977) have have shown
shown that,that, within
within the the framework
framework of of cur-
cur-
rent standard
rent standard cosmology
cosmology (Robertsonian
(Robertsonian models), models), all all possible
possible observations
observations
fail to
fail to univocally
univocally determine
determine the the global
global topology
topology of of spacetime.
spacetime. This This doesn't
doesn’t
preclude that local models have to satisfy PS. In fact, we know that in rather
preclude that local models have to satisfy PS. In fact, we know that in rather
large regions
large regions of of the
the universe
universe we we can
can safely
safely useuse thethe special
special theory
theory of of relativity
relativity
provided the
provided the phenomena
phenomena studied studied areare not
not too
too sensitive
sensitive to to spacetime
spacetime curvature.
curvature.
We also
We also know
know that the invariance
that the invariance group group of of the
the general
general theory
theory of of relativity,
relativity, the the
group of
group of diffeomorphisms,
diffeomorphisms, is is wider
wider than
than thethe Lorentz-Poincaré
Lorentz-Poincaré group. group. How- How-
ever, this
ever, this doesn't
doesn’t prevent
prevent us us to
to embed
embed local
local phenomena
phenomena in in Minkowski
Minkowski space- space-
time as is guaranteed by the Principle of equivalence. In other words,
time as is guaranteed by the Principle of equivalence. In other words, locally,
locally,
the symmetry
the symmetry principle
principle holds.
holds. And And thethe local
local validity
validity of the special
of the special theory
theory of of
relativity imposes
relativity imposes constraints
constraints on on the
the kind
kind of of metric
metric which
which is is globally accept-
globally accept-
able. More precisely
able. More precisely the the metric
metric mustmust bebe Riemannian,
Riemannian, 4-dimensional
4-dimensional and and of of
negative signature
negative signature (See (See Brown
Brown 1997).1997).
Of course, the
Of course, the possibility
possibility of of giving
giving alternative
alternative and and incompatible
incompatible metaphys- metaphys-
ical interpretations of the same theory
ical interpretations of the same theory remains open. Forremains open. For example,
example, we we could
could
ask ourselves if
ask ourselves if there
there exists some sort
exists some sort ofof substantival
substantival spacetime
spacetime or or ifif space-
space-
time points
time points areare individuals.
individuals. Such Such moves
moves are are neither
neither allowed
allowed nor nor forbidden
forbidden
by the
by the brand
brand of of moderate
moderate and and selective
selective scientific
scientific realism
realism advocated
advocated here here (see(see
also Ghins 1992),
also Ghins 1992), whichwhich confines
confines its its existential
existential claimsclaims to to physical
physical objects
objects
 VAN FRAASSEN’S
VAN FR CONSTRUCTIVE
AASSEN 'S C EMPIRICISM
ON STR U C TI VE EM PI RI CI SM 3 4 1 341

only to
only to the
the extent
extent that
that they
they are
are represented
represented by
by invariant
invariant mathematical
mathematical ob-
ob-
jects in the context of empirical substructures of models which, on
jects in the context of empirical substructures o f models which, on top of top of
satisfying the symmetry principle PS, are also empirically adequate.
satisfying the symmetry principle PS, are also empirically adequate.

Centre de philosophie
Centre de des sciences
philosophie des sciences
Institut
Institut Supérieur de Philosophie
Supérieur de Philosophie
Place du cardinal Mercier, 14
B-1348 Louvain-la-Neuve, Belgium
B-1348 Louvain-la-Neuve, Belgium
E-mail:
E [email protected]
-
REFERENCES
REFERENCES
m
a
BERZI, V.
BERZI, V. and
and GORINI,
GORINI, V. V. (1969),
(1969), "Reciprocity
“Reciprocity
i Principle and
Principle and the
the Lorentz
Lorentz
Transformations”, J.
Transformations", J. ooff Math. Physics
Math. Physicsl 10, 8, 1518–24.
10,8,1518-24.
BROWN, H.
BROWN, H. R.
R. (1997),
(1997), "On
“On the the Role
Role :of
of Special Relativity in
Special Relativity in General
General Rela-
Rela-
tivity”, International
tivity", International Studies
Studies in in the
the Philosophy of
g Philosophy of Science,
Science, 67–81.
67-81.
FRIEDMAN, M.
FRIEDMAN, M. (1983),
(1983), Foundations
Foundations h o off Space-Time
Space-Time Theories.
Theories. Princeton
Princeton
University Press.
University Press. i
GEMIGNANI,
GEMIGNANI, M. M. C.C. (1971),
(1971), Axiomatic
Axiomatic n Geometry,
Geometry, Addison-Wesley.
Addison-Wesley.
GHINS,
GHINS, M. (1992), “Scientific
M. (1992), Realism
"Scientific Realism and Invariance”.
s and Invariance". Proceedings
Proceedings of of the
the
Third
Third SOFIA
SOFIA Conference
Conference on on Epistemology.
@
Epistemology. Campinas. July
Campinas. July 30–August
30—August1, 1,
1990. Philosophical Issues
1990. Philosophical Issues (Vol.
(Vol. 2:
2: Rationality in
r Rationality Epistemology). 249—
in Epistemology). 249–
62. California:
62. California: Ridgeview.
Ridgeview. i
GHINS,
GHINS, M. M. and
and BUDDEN,
BUDDEN, T. T. (2001),
(2001),s“The
"The Principle
Principle of
of Equivalence”.
Equivalence". Stud-
Stud-
ies in
ies in History
History and
and Philosophy
Philosophy of of Modern
p
Modern Physics. Vol.
Physics. Vol. 32B,
32B, n.n. 1.
1. March
March
2001, 33–51.
2001,33-51. .
GLYMOUR,
GLYMOUR, C. C. (1977),
(1977), “Indistinguishable
"Indistinguishable u Space-Times
Space-Times and and the
the Fundamen-
Fundamen-
tal Group"
tal Group” in in J.
J. Earman,
Earman, C. C. Glymour,
Glymour,
c J. Stachel.
J. Stachel. Foundations
Foundations of of Space-
Space-
Time Theories. Minnesota
Time Theories. Minnesota StudiesStudiesi inin the
the Philosophy
Philosophy of o f Science. Vol.
Science. Vol.
VIII. Minneapolis. 50–60.
VIII. Minneapolis. 50-60. .
KANT,
KANT, I. I. (1970),
(1970), Metaphysical
Metaphysical Foundations a
Foundations of
of Natural
Natural Science. (Transl. by
Science. (Transi. by
J. Ellington).
J. Ellington). Indianapolis.
Indianapolis. c
MALAMENT,
MAL AMENT, D. D. (1977),
(1977), "Observationally
“Observationally
. Indistinguishable Space-times:
Indistinguishable Space-times:
Comments on
Comments on Glymour’s
Glymour's Paper" Paper” binin J.
J. Earman,
Earman, C. C. Glymour,
Glymour, J. J. Stachel.
Stachel.
Foundations of
Foundations of Space-Time Theories.
Space-Time Theories. Minnesota Studies
e Minnesota Studies in the Philoso-
in the Philoso-
phy of
phy of Science. Vol. VIII.
Science. Vol. VIII. Minneapolis.
Minneapolis. 61-80.
61–80.
McMULLIN,
McMUL E. (1984),
L IN, E. (1984), "A“A Case
Case for
for Scientific
Scientific Realism"
Realism” in
in Leplin,
Leplin, J.
J. (ed.)
(ed.)
Scientific Realism. Berkeley: University of California Press.
Scientific Realism. Berkeley: University of California Press.
REICHENBACH, H.
REICHENBACH, H. (1958),
(1958), The
The Philosophy
Philosophy of of Space
Space and
and Time. New York:
Time. New York:
Dover.
SUPPE, F. (1989),
SUPPE, F. (1989), The
The Semantic
Semantic Conception
Conception of o f Theories
Theories and
and Scientific
Scientific Re-
Re-
alism. University of Illinois Press.
alism. University of Illinois Press.
342 M I C H E MICHEL
L GHINS
GHI NS

VAN FRAASSEN,
VAN FRAASSEN, B. B. (1980),
(1980), The
The Scientific Image. Oxford
Scientific Image. Oxford University
University Press.
Press.
VAN FRAASSEN, B. (1989), Laws and Symmetry. Oxford University
VAN FRAASSEN, B. (1989), Laws and Symmetry. Oxford University Press. Press.
WEYL, H.
WEYL, H. (1963),
(1963), Philosophy
Philosophy of Mathematics and Natural Science. New
York:
York: Atheneum.
Atheneum.