Escolar Documentos
Profissional Documentos
Cultura Documentos
DOI 10.1007/s10670-013-9578-5
ORIGINAL ARTICLE
Sebastian Lutz
1 Introduction
The analysis of scientific theories needs some frameworka way in which theories
and possibly the world are described and, building on that, a set of tools for analysis.
The earliest formal framework, developed within logical empiricism and especially
by Carnap and Hempel, has been dubbed the Received View. It relies on
formalizations of scientific theories in languages of predicate logic and assumes that
the non-logical vocabulary is bipartitioned into observational and theoretical terms,
where only the observational terms are directly interpreted. The interpretation of the
theoretical terms is fixed only by the interpretation of the observation terms, the
S. Lutz (&)
Munich Center for Mathematical Philosophy, Ludwig-Maximilians-Universitat Munchen,
Geschwister-Scholl-Platz 1, 80539 Munchen, Germany
e-mail: sebastian.lutz@gmx.net
123
S. Lutz
1
I will use approach and framework interchangeably in the following.
2
Although this categorization of syntactic and semantic approaches is standard terminology, it is
somewhat incongruous: Set theory and predicate logic can both be used as foundational languages in
which to formalize other theories. Model theory, on the other hand, is one of those theories that can be so
formalized. I will discuss this point further at the end of 4.
3
The question what scientific theories really are may also simply not be well-defined, so that there is no
fact of the matter.
123
Whats Right with a Syntactic Approach?
R has a fixed vocabulary V, containing mi-place predicates Pi, nj-place functions Fj,
constants ck, and in higher order logic their respective types. V is contained in the
image of U as well, since every structure S 2 S contains a mapping from every
element of V to an extension with the corresponding arity and type (Hodges 1993,
24; Chang and Keisler 1990, 1.3). Call the arities and types of the extensions the
similarity type of V. V and its similarity type, sometimes called the signature of S,
can thus be read off uniquely from S (Hodges 1993, 4).4
However, U does lose some information because it cannot distinguish between
equivalent sets of sentences, that is, if R H, then UR UH. This can pose
problems, for example when modifying a theory: One formulation of a theory can
be vastly superior to an equivalent one when it needs to be generalized or adjusted
(van Fraassen 1980, 3.5). Relatedly, the formulation is also relevant when it comes
to the inductive support of components of the theory: If the data support one
postulate but not another, a formulation that keeps the two postulates separate is
arguably better than one that contains a single postulate equivalent to their
conjunction. The loss of distinction between equivalent sets of sentences does not
pose a problem, however, if the results of an analysis of a scientific theory are
invariant under the theorys equivalent reformulation; and outside of questions of
induction, many interesting analyses of scientific theories are so invariant. Indeed, it
is often considered a problem if an analysis of the theory is not (witness, for
example, Carnap 1956, 56).
Conversely, any set of structures S yields a set of sentences R through the
mapping
4
This is discussed at length in Sect. 3.
5
Note that this is a weaker condition than the point-wise isomorphism defined by Halvorson (2012, 190).
123
S. Lutz
6
See, for example, van Fraassens notion of a theory (van Fraassen 2008, 238).
123
Whats Right with a Syntactic Approach?
3 Language Independence
According to van Fraassen (1980, 44), the same class of structures could well be
described in radically different ways, each with its own limitations, and in the
words of Suppe (1989, 4), a semantic approach construes theories as what their
formulations refer to when the formulations are given a (formal) semantic
interpretation. French and Ladyman (1999, 114115) similarly assume that the
structures used in semantic approaches do not contain a vocabulary when they
discuss a criticism of semantic approaches they attribute to Mauricio Suarez: If a
semantic approach uses models as they are defined in model theory, it is still
dependent on a language, since a model is a structure and an interpretation of a
formal language in terms of that structure (that is, a map from the symbols of the
syntax to elements of the structure). If models are taken to involve such a mapping,
French and Ladyman (1999, 114) write, it is clear that the celebrated claim of the
linguistic independence of considering models (and not first-order formalizations of
theories), stressed by adherents of the semantic approach as giving it a clear
advantage over the syntactic view, is simply not true. A similar position is taken by
van Fraassen (1989, 366), and French and Ladyman (1999, 115) conclude that van
Fraassen should be interpreted as talking about structures by those who wish to
understand model in the sense of the standard logic texts.
However, it is not exactly clear what French and Ladyman mean by structure,
except that a structure contains the extensions of symbols, and not the symbols
themselves. They do claim that van Fraassens emphasis on structure is compatible
with this definition of model theory from a contemporary textbook, according to
which model theory is the study of the construction and classification of structures
within specified classes of structures. Of course, everything in this quote from
Hodges (1993, ix) depends on his definition of structure. Here is, for example, the
part of his definition that deals with relations (Hodges 1993, 2, my notation):
For each positive integer n [a structure contains] a set of n-ary relations on jAj
(i. e. subsets of jAjn), each of which is named by one or more n-ary relation
symbols. If R is a relation symbol, we write RA for the relation named by R.
It is clear that, as assumed so far and contrary to French and Ladymans claim, the
symbols play an important role in Hodges definition of a structure: They identify
the extensions by naming them.7
A definition better suited to French and Ladymans position can be found in a
less contemporary textbook (Bell and Slomson 1974, 3.2), in which
7
Indeed, Hodges (1993, 12) explicitly agrees with this assessment.
123
S. Lutz
8
It is clear from their use of as an index set that Bell and Slomson (1974) intend fRn jn 2 xg to be an
indexed set, in my notation fRn gn2x (rather than the set of the indexed sets elements). For if R2 = R3,
then fRi gi2f1;2;3g 6 fRi gi2f1;2g ; while fRi ji 2 f1; 2; 3gg fRi ji 2 f1; 2gg; Bell and Slomson thus need
indexed sets to allow for different names with identical extension.
123
Whats Right with a Syntactic Approach?
the extension of Q0 (which is possible because the extensions are not ordered), so
that the model is a model of K as well. Thus every model of H is a model of K and
vice versa. Therefore, if the extensions in a structure were not ordered, semantic
approaches in principle could not distinguish all those theories that can be
distinguished syntactically.
Hence for the definition of a variant of the notion of a structure that specifies no
vocabulary, something like a tuple is needed. But since structures can have infinite
vocabularies and tuples are finite, such a variant cannot be defined in terms of tuples if
it is to be able to express anything that can be expressed in terms of structures. In line
with the possibility of assigning any finite well-ordered index set to tuples, I therefore
suggest defining pure structures so that they contain mappings from arbitrary and not
fixed well-ordered index sets to extensions. That the index sets are not fixed can be
captured by identifying any two mappings that differ only in their index sets:
Definition 1 A representative of a pure structure A ^ is a triple hA; a; i; where
a : I ! I V is a mapping from the index set I to the image of an interpretation I ,
and is a well-ordering of I. Two triples hA; a; i and hA; a0 ; 0 i represent the same
pure structure, hA; a; i hA; a0 ; 0 i, if and only if there is an order isomorphism
f : I ! I 0 and a0 f a.
The use of an interpretation I is just a convenient, non-essential way of ensuring
that a maps only to set theoretical objects that can be extensions of predicate,
function, or constant symbols. The definition determines a tuple if and only if the
index sets of the representatives are finite.
Definition 1 does not introduce a specific vocabulary, and is as language-
independent as the use of tuples. To stress the point: The infinity of representatives
with different vocabularies in definition 1 is not artificially introduced. Both indexed
sets and tuples assume some specific set or multiple sets of entities that provide a
vocabulary. Index sets can be used as vocabularies immediately, since they are
already mappings from some set to a set of extensions. Tuples allow the
introduction of such a mapping without any further assumptions except about
the elements of the index set. In definition 1, this further assumption is avoided by
the identification of representatives with different index sets.
The indexed structures given by Bell and Slomson (1974) relate to structures in a
very simple way: Their index set I plays the role of the set of relation symbols
fPi ji 2 Ig of structures.9 Definition 1 requires a somewhat more elaborate
modification of standard definitions, of which I will only give the modification
for the notion of embedding. Since pure structures are given by classes of mappings
with well-ordered index sets, embeddings between pure structures can be defined
via a bijection between the index sets of their representatives and a function from
one pure structures domain to the others. In this definition (which I give only for
pure first order structures), the position in the ordering of the index sets plays the
role of the element of the vocabulary.
9
For examples relevant in the following, see the definitions of reduct, isomorphism, and
substructure by Bell and Slomson (1974, 153, 73) and by Chang and Keisler (1990, 20-23),
respectively. The generalization to structures with function and constant symbols is straightforward.
123
S. Lutz
Definition 2 A pure rst order structure A ^ can be embedded in a pure first order
^
structure B if and only if for any two representatives A ^ hA; a; i with
^
a : I ! I V and an ordering on I, and B hB; b; i with b : J ! J V 0
0
123
Whats Right with a Syntactic Approach?
^
^ hA; a; i hA; c; i; B
Fig. 2 Functions between four representatives of two structures A
0
hB; b; 0 i hB; d; i so that A ^
^ can be embedded in B
Conversely, a structure A can be embedded in a structure B if and only if, under some
^ respectively,
^ and B,
well-ordering of their vocabulary, they represent pure structures A
^ can be embedded in B.
such that A ^ The proof assumes the axiom of choice.
123
S. Lutz
123
Whats Right with a Syntactic Approach?
However, semantic approaches are too liberal in their neutrality with respect to
the vocabulary, for not every renaming makes sense. In the ideal gas law, for
example, the translation should allow the renaming of temperature into
temperatur, but not into pression. For syntactic approaches, this restriction on
renaming can be expressed by introducing analytic sentences. Any renaming not
entailed by analytic sentences is not allowed. Then, temperature can be renamed
as temperatur, but not as pression.10 To avoid such incorrect renaming in
semantic approaches, the connection between pure structures and the world has to
be explicitly taken into account (see Sect. 4).
There is another problem with arbitrary renaming, for it is not only too liberal in
some respects, but also too restrictive in others. It does not, for example, avoid the
language dependence referred to by Suppe (1974, 204205), who notes that wave
and matrix mechanics, although the same theory, constitute different sets of
sentences.11 Obviously, the difference between matrix mechanics and wave
mechanics goes beyond a mere renaming of the predicate, function, and object
symbols. And this is a problem for semantic approaches, because if not only the
names of the extensions change, the pure structures change.
To accommodate the possibility of restricted renaming, I have used the simplest
kind of definition, an identification of symbols. But the comparison of two theories
can be generalized to include any kind of explicit definitions, and this generalization
can accommodate the restricted change of structure. Two theories H; H0 are
denitionally equivalent if and only if both can be extended by explicit definitions
such that they become equivalent. Then their models can be turned into each other
by a procedure analogous to the one described above: First expand the model of one
theory to include the defined symbols of the definitions. This expansion is unique.
Then reduce the resulting structure to the vocabulary of the other theory (cf. Hodges
1993, 61). This allows the identification of theories that differ not only in the
vocabulary they use, but also in the pure structures that their models represent. The
procedure therefore also allows for structural differences in the description of
theories, and thus can be used to extend not only syntactic approaches, but also
semantic ones. It leads to an equivalence class of classes of structures that are taken
to formalize the same theory. This move would also not obviously tether semantic
approaches further to syntactic ones because the notion of definitional equivalence
can be defined without reference to sets of sentences (de Bouvere 1965).
A set R of sentences of predicate logic is not enough for applying a theory to the
world, for if A R, any set of the same cardinality as jAj can be made into a model
of R as well:
10
Incidentally, if one wanted to identify theories with sets of propositions, the identification of
expressions for the same proposition would provide analytic sentences.
11
Halvorson (2012, 4.2) gives much more precise examples.
123
S. Lutz
each tuple hb1 ; . . .; bnj i 2 Bnj (or arguments for functions of higher order), define
fjB b1 ; . . .; bnj : gfjA g1 b1 ; . . .; g1 bnj , and for each constant cA
k , define
cBk : gc A
k . It is straightforward to show that B A: h
The bijection g may in particular be a permutation on a domain, so that any
element of A can be exchanged for any other element.
Claim 7 Let A R and cardjAj = card(B). Then there is a B with jBj B and
B R.
Proof (Przecki 1969, 2728) If cardjAj = card(B), then there exists a bijection
g : jAj ! B. The corollary follows from lemma 6, because sets of sentences can
determine structures at most up to isomorphism. h
In light of claim 7, it is clear that a set of sentences can be connected to the world
(beyond statements about the number of its objects) only if there is a means of
distinguishing between isomorphic structures. One way of solving this problem is
by introducing a class of possible structures M, which is determined by the
extensions of the V-terms given their meaning.12 M allows for R making more than
cardinality claims about the world by defining the possible models of R as the
possible structures that are models of R. The use of possible structures does not have
to trivialize the distinction between semantic and syntactic approaches: One can
postulate that M only distinguish between isomorphic structures, so that the
isomorphic closure of M is always the class of all V-structures. Then it is easy to
show that the isomorphic closure of the possible models of R is always the class of
all models of Rin other words, in syntactic approaches, non-isomorphic structures
have to be distinguished by sentences of the object language. This suggests that, for
the purposes of this discussion, syntactic approaches are best conceptualized as
follows:
Definition 3 Syntactic approaches describe theories and their relations to the
world using sentences in an object language and structures. Non-isomorphic
structures are only distinguished by sentences in the object language.
12
Note that the domains of the structures in M do not generally contain only actual objects (Lutz 2012a,
188).
123
Whats Right with a Syntactic Approach?
How the connection between pure structures and the world is to be envisaged in
semantic approaches depends on whether the pure structures are meant to be
worldly or non-worldly. Suppes statement quoted at the beginning of Sect. 3 (that
semantic approaches construe theories as the formal referents of the theories
formulations) suggests that the pure structures used in semantic approaches contain
worldly extensions, that is, are worldly themselves. Da Costa and French (2000, fn.
2) also seem to assume that the pure structures are worldly when they state that the
set-theoretic models are constructed in set theories with Urelemente (individual[s],
systems, portions of the universe, real things,).
Representatives of a pure structure are structures, and accordingly the existence
of a worldly pure structure entails the existence of a worldly structure. This structure
then provides the connection of the linguistic entities with the world. Therefore, if a
123
S. Lutz
13
For indexed structures, the modification of this argument is straightforward: Each structures index set
I can be used directly as a vocabulary, providing the interpretation a : I ! aI. If a specific vocabulary
V is desirable, any bijection g : V ! I leads to the interpretation I a g : V ! aI. The discussions
below can modified analogously.
14
In semantic approaches, isomorphisms are often used as analogues to Tarskis definition of truth.
15
Muller presents his formalism as an extension of his notion of structures, in contradiction to the
conclusion reached in connection with definition 1.
123
Whats Right with a Syntactic Approach?
represent real events and processes cannot be determined by the content of the
theory, but is a pragmatic fact about our language [] and it is unreasonable
to demand that the semantic view explains the nature of representation in
general.
This clearly falls short of the idea that semantic approaches are easier to connect to
the world than syntactic approaches.
In French and Ladymans approach, structures are simply set theoretic constructs,
and thus in this respect their view is identical to that of the structuralists following
Sneed (1971) and Stegmuller (1979), whose formal core commitment is to the
expression of scientific theories in terms of set theory. In such a structuralist
approach, there is simply one set of set theoretic sentences (those describing the
phenomena) that are distinguished as representing real events. But if neither
predicate logic nor set theory are given a formal connection to the world, they are
simply languages that are assumed to describe the world in some not further
specified way. Then, for example, the question of non-standard models does not
even arise (neither for first order nor higher order logic, nor set theory) unless an
additional metalanguage is artificially introduced (Vaananen 2001). And if there is a
metalanguage, set theory and higher order logic are equally expressive (Vaananen
2001, 506507). The discussion in Sect. 2 about the possibilities to capture
structures up to isomorphism therefore stacks the deck against syntactic approaches:
It rests on the heroic assumption that set theory is not a language that describes the
world, but is the world (more precisely: the world is a worldly structure), and the
task of syntactic approaches is to describe the set theoretic world. If one instead
treats, with French and Ladyman, set theory as just another language (which it is), it
cannot describe the world more precisely than higher order logic.16
If the relation to the world is excluded from the formalisms of syntactic and semantic
approaches, then the difference between the two reduces to the question which syntax
(and which axioms) to use in formalizations. Semantic approaches assume the axioms of
(possibly nave) set theory and often assume that all further (mathematical) concepts
have been introduced by explicit definitions. Hilbert (1900, 10921093) calls this the
genetic method of formalization. Syntactic approaches instead rely on predicate
logic. In principle, syntactic approaches could then follow the genetic method and first
formalize set theory, then introduce further concepts by explicit definitions. But
syntactic approaches can also follow Hilberts axiomatic method and only provide the
axioms for those terms that actually appear in the theory that is to be formalized, without
going through set theory first. Therefore, semantic approaches are a subclass of syntactic
ones if the relation to the world is excluded from the formalisms.
In conclusion, the connection to the world is equally problematic for syntactic
and semantic approaches. And a solution for one kind of approach also provides a
solution for the other.17
16
Vaananen (2001) further argues that neither set theory nor higher order logic can describe the world
more precisely than first order logic.
17
I have argued that solutions for semantic approaches are also solutions for syntactic approaches. The
argument for opposite direction can rely on the fact that every structure represents a pure structure and, if
needed, corollary 5.
123
S. Lutz
5 Corollaries
The remaining criticisms of syntactic approaches, that they are too cumbersome,
cannot capture scientific models, and cannot capture the relation between theory and
observations, had been developed for the Received View and were then leveled
against all syntactic approaches without further justification. The following
discussion can hence be kept short.
That formalizations in the Received View are too cumbersome has been inferred
from the (incorrect) assumptions that these formalizations have to be given in first
order logic and have to explicitly contain all axioms for all terms that occur in a
theory (Lutz 2012c, 2, 3). Syntactic approaches in general, however, can clearly
avail themselves to higher order logic, and second order logic is already enough to
capture directly most all mathematical practice (Leivant 1994, 260; cf. Vaananen
2001, 515).18 Since capturing mathematical practice only becomes easier for higher
orders, reliance on predicate logic does not lead to cumbersome formalizations. The
possible counterargument that higher order logics do not have a complete proof
theory has been addressed in Sect. 2. Syntactic approaches can clearly also leave
some axioms implicit, to be spelled out in detail when needed. Conversely, the
axioms of set theory and model theory can also be explicitly spelled out, and so can
the additional definitions and axioms for further mathematical concepts (cf. Lutz
2012c, 2).
The claim that the Received View is less hospitable to scientific models than
semantic approaches stems from a misunderstanding of some statements by
proponents of the Received View (Lutz 2012c, 4) and, it seems, the identification
of models in the model theoretic sense with scientific models. The former point is
irrelevant for syntactic approaches in general; the latter point is wrong because the
different meanings and uses of the term model in the natural sciences clearly do
not fit the definition of model in model theory.19 Without this identification, the
question becomes whether semantic approaches can describe the world (specifically,
the scientific models in the world) better than syntactic approaches. I have answered
this question in the negative in Sects. 2 and 4.
After an analysis of the Received Views reliance on correspondence rules and a
bipartitioned vocabulary, Suppe (1974, 114) concludes that if formalization is
desirable in a philosophical analysis of theories, it must be of a semantic sort. But
syntactic approaches do not need to assume a bipartition of the vocabulary, and
hence also do not need to assume a connection of theory and observations by
correspondence rules. And not only is Suppes argument invalid, Suppe himself
shows that his conclusion is also wrong: His criticism of correspondence rules
(Suppe 1974, II.E) is based on a syntactic description of a relation between theory
and observations that he takes to be more adequate (cf. Suppe 1974, 108, n. 225).
This trivially establishes the possibility of capturing the relation syntactically.
18
Andrews (2002, xii-xiii) even argues that higher order logic provides a more natural formalization of
mathematical reasoning than set theory.
19
The influential argument to the contrary by Suppes (1960) establishes at best that the formal part of
scientific models can, with reformulations, be expressed in set theory (cf. Lutz 2012c, 9799).
123
Whats Right with a Syntactic Approach?
6 On Peaceful Coexistence
The most important result of the preceding discussion is the close connection between
syntactic and semantic approaches. It is easy to go from pure structures to indexed
structures, to structures, to sentences, and back again. Accordingly, it is easy to move
from a semantic approach to a syntactic one and back. On the one hand, that means that
the problems of syntactic approaches reappear in semantic approaches, as the
problems of connecting the respective formalisms to the world show. On the other
hand, it means that the solutions found in syntactic approaches can be used in semantic
approaches as well, for example the use of explicit definitions for the identification of
sets of sentences and classes of structures.
I hope that in the future more solutions from one kind of approach will find their
ways to the other. For on a meta-level, I do hold the position of the critics of
syntactic approaches: The language in which an analysis is phrased, whether it uses
pure or indexed structures, structures, or possible structures and an object language,
often matters very little.
Acknowledgments A very early version of this section has been presented at Herman Philipses Dutch
Research Seminar in Analytic Philosophy at Utrecht University. Parts have been presented at the EPSA 09
at the Vrije Universiteit, Amsterdam, on October 23, 2009 and at the workshop Perspectives on
Structuralism at the Center for Advanced Studies/ Munich Center for Mathematical Philosophy, Ludwig-
Maximilians-Universitat Munchen, on February 17, 2012. I thank the participants for helpful discussions.
This article has also profited a lot from a reading group at Tilburg University with Reinhard Muskens and
Stefan Wintein. It is based on 4.1 of my dissertation prepared at Utrecht University (Lutz 2012b), which
contains elaborations of some of the claims of this article and additional references. I thank my advisers
Janneke van Lith, Albert Visser, and especially Thomas Muller for extensive and helpful comments and
discussions, and F. A. Muller for his trenchant comments on my dissertation in general and the themes of
this article in particular. I thank Arno Bastenhof for helpful discussions, an anonymous reviewer for
helpful comments, and Alana Yu for helpful suggestions in matters of style. Parts of the research for this
article have been supported by the Alexander von Humboldt Foundation.
References
Andrews, P.B. (2002). An introduction to mathematical logic and type theory: To truth through proof (2nd
ed.). Berlin: Springer.
Bell, J. L., & Slomson, A. B. (1974). Models and ultraproducts: An introduction (3rd ed.). Amsterdam:
North-Holland.
Carnap, R. (1956). The methodological character of theoretical concepts. In H. Feigl & M. Scriven (Eds.),
The foundations of science and the concepts of psychology and psychoanalysis. Volume 1 of
Minnesota studies in the philosophy of science. Minneapolis, MN: University of Minnesota Press.
Carnap, R. (1939). Foundations of logic and mathematics. Volume I, Issue 3 of Foundations of the Unity
of Science. Toward an International Encyclopedia of Unified Science. Chicago, London: University
of Chicago Press. References are to the two-volume edition.
Chakravartty, A. (2001). The semantic or model-theoretic view of theories and scientific realism.
Synthese, 127, 325345.
123
S. Lutz
Chang, C. C., & Keisler, H. J. (1990). Model theory. Volume 73 of Studies in logic and the foundations of
mathematics (3rd ed.). Amsterdam: North Holland. 3rd impression 1992.
da Costa, N., & French, S. (1990). The model-theoretic approach in the philosophy of science. Philosophy
of Science, 57, 248265.
da Costa, N., & French, S. (2000). Models, theories, and structures: Thirty years on. Philosophy of
Science, 67(Proceedings), S116S127.
de Bouvere, K. (1965). Synonymous theories. In: J. W. Addison, L. Henkin, & A. Tarski (eds.), The
theory of models. Proceedings of the 1963 international symposium at Berkeley, (pp. 402406)
Amsterdam: North-Holland Publishing.
Enderton, H. B. (2009). Second-order and higher-order logic. In: E. N. Zalta (Ed.), The Stanford
encyclopedia of philosophy. Center for the study of language and information. Stanford University,
Spring 2009 edition.
van Fraassen, B.C. (1980). The scientic image. Oxford: Clarendon Press.
van Fraassen, B. C. (1989). Laws and symmetry. Oxford: Clarendon Press.
van Fraassen, B.C. (2008). Scientic representation: Paradoxes of perspective. Oxford: Clarendon Press.
French, S., & Ladyman, J. (1999). Reinflating the semantic approach. International Studies in the
Philosophy of Science, 13(2), 103121.
Halvorson, H. (2012). What scientific theories could not be. Philosophy of Science, 79(2), 183206.
Hilbert, D. (1900). Uber den Zahlenbegriff. In G. Hauck, & A. Gutzmer (Eds.), Jahresbericht der
deutschen Mathematiker-Vereinigung (Vol. 8, pp. 180183). Leipzig. B. G. Teubner. References are
to the translation (Hilbert, 1996).
Hilbert, D. (1996). On the concept of number. In: W. Ewald (Ed.), From Kant to Hilbert: A source book in
the foundations of mathematics, volume II. (pp. 10921095). Clarendon Press, Oxford. Digitally
reprinted in 2005.
Hodges, W. (1993). Model theory. Cambridge: Cambridge University Press. Digitally printed in 2008.
Ketland, J. (2004). Empirical adequacy and ramsification. The British Journal for the Philosophy of
Science,, 55, 287300.
Kitcher, P. & Salmon, W., (eds) (1989). Scientic explanation. Volume 13 of Minnesota studies in the
philosophy of science. Minneapolis, MN: University of Minnesota Press.
Leivant, D. (1994). Higher order logic. In D. M. Gaby, C. Hogger & J. Robinson (Eds.), Deduction
methodologies. Volume 2 of Handbook of logic in artificial intelligence and logic programming (pp.
229321). Oxford: Oxford University Press.
Lutz, S. (2012a). Artificial language philosophy of science. European Journal for Philosophy of Science, 2(2),
181203.
Lutz, S. (2012b). Criteria of empirical significance: Foundations, relations, applications. PhD thesis,
Utrecht University. http://philsci-archive.pitt.edu/id/eprint/9117.
Lutz, S. (2012c). On a straw man in the philosophy of science: A defense of the Received View. HOPOS:
The Journal of the International Society for the History of Philosophy of Science, 2(1), 77120.
Muller, F. A. (2010). Reflections on the revolution at Stanford. Synthese, 183(1), 87114.
Newman, M. H. A. (1928). Mr. Russells causal theory of perception. Mind, 37(146), 137148.
Przeecki, M. (1969). The logic of empirical theories. London, New York: Routledge & Kegan Paul/
Humanities Press.
Sneed, J. D. (1971). The logical structure of mathematical physics. Dordrecht, The Netherlands: D. Reidel
Publishing Co..
Stegmuller, W. (1979). The structuralist view of theories. New York: Springer.
Suppe, F. (1974). The search for philosophic understanding of scientific theories. In F. Suppe (Ed.), The
structure of scientic theories (pp. 3241). Urbana, IL: University of Illinois Press.
Suppe, F. (1989). The semantic conception of theories and scientic realism. Urbana, IL: University of
Illinois Press.
Suppe, F. (2000). Understanding scientific theories: An assessment of developments, 19691998.
Philosophy of Science, 67, S102S115.
Suppes, P. (1960). A comparison of the meaning and uses of models in mathematics and the empirical
sciences. Synthese, 12, 287301.
Suppes, P. (2002). Representation and invariance of scientic structures. Stanford, CA: CSLI
Publications.
Vaananen, J. (2001). Second-order logic and foundations of mathematics. The Bulletin of Symbolic Logic,
7(4), 504520.
123