Fritz Machlup as a precursor of the Semantic Conception in Philosophy of Science rbeltramino@ucel.edu.ar 1.

Introduction Few authors in the Austrian Tradition are more frequently overlooked as Fritz Machlup, probably his proverbial Viennese politeness and his intellectual moderation contributed to this. Nowadays he remains quite in the shadow of more known authors like Mises or Hayek, although in the particular field of Philosophy of Science, he was by far one of the most important authors coming from the Austrian Tradition into the mainstream. His discussions with Terence W. Hutchison about empirism and verification and with Richard Lester about firms theory premises are still studied and quoted widely and his position is still considered an interesting alternative to what he once labelled ultraempirism. Machlup was through all his life, very interested and very up to date with the developments and different currents in Philosophy of Science 1. So when he wrote about economic methodology he was a rara avis, a successful practitioner who was also well read in Philosophy of Science. In that, he followed his own rule, which he clearly stated as a piece of advice to every economist:

I often tell my students that they should not publish any methodological notes, papers or books until they have done years of substantive research in their field and attained recognition for their mastery of its technical aspects (Machlup 1978 p.X)
Unfortunately the development of the Methodology (with capital M, notwithstanding Mc Closkeys criticism of it) of Economics as a separate field made that piece of advice obsolete. In our time is practically impossible to join the two characters, the successful practitioner and the well read philosopher in one individual. Machlup could and that is the reason why is so interesting to rightfully understand his position. There is so much sheer ignorance of Philosophy of Science in nearly every economist writing about Philosophy of Science 2 that Machlups work stands out from the group. Machlup himself felt necessary to explain his strong taste for methodology and he did so, pointing to the Viennese intellectual milieu3 where he grew up as the reason:

I have often been asked how I developed such a strong taste for methodological discourse. The answer is easy: in the intellectual milieu in which I lived it would have been surprising for any student to remain uninterested in methodology. (Machlup 1978 p.I)

1 2

Most of his articles in this matter are included in Machlup (1978) On the opposite not many philosophers of science wrote about economics. Caldwell (1994) shows the problems and confusions that this economists lack of knowledge while writing about Philosophy of Science had caused through the twentieth century. 3 The classic book about that irrepetible viennese millieu is Janik and Toulmin (1973) and his sequel Janik (2001)

Then, the Austrian author gives an impressive list of some of the most relevant individuals in that milieu. The list is a whos who of twentieth century intellectuals and consists of:
Menger, Mach, Weber, Schlick, Waismann, Carnap, Neurath, Hahn, Feigl, Wittgenstein, Popper, Gdel, Mises, Kaufmann y Schtz. I have not omitted anyone, because Machlup didnt. Its for Machlups strong taste in these matters that I think is particularly interesting to read and understand Machlups position in the light of some of the most promising developments in Philosophy of Science and make a critical comparison of his work and this recent developments. As Caldwell (1982) says: it is still clear that Machlup had a better understanding of the philosophy of science that was contemporaneous with his efforts than did Hutchison and Machlups methodological schooling allowed to formulate a position consistent with the philosophy of science of his time (Caldwell 1982 p.145) The main danger of the kind of reinterpretation that is here proposed is always to fall prey of some empty scholasticism. To avoid it, the emphasis must be put in the actual practitioners of Economics because we wish to interrogate Machlup about his methodological position not for Machlups memory sake4. This work intends to show that Machlups well known and influential position in Economic Methodology can be better understood in the frame of the Semantic Conception, therefore allowing a better elucidation of economists practice of science for Philosophy of Science. In the following section (II) I will briefly analyze Machlups position through primary and secondary sources, then in section III I will again briefly but sufficiently for the purpose of this article characterize the Semantic Conception in Philosophy of Science. Later in section IV, I will try to show the coincidences and differences between them and finally in section V, I will extract some conclusions for future work.

II. Fritz Machlup as a Lakatosian precursor The idea of understanding Machlups work as a precursor of Lakatos came to me when I was introduced to Lakatoss work in graduate school. I thought I knew Machlups work and Lakatosconception of science appeared strikingly similar. Apparently I had found my thesis theme. Later I discovered that an Argentinian author Gabriel Zanotti had not only thought, but written and published something very similar5 several years before. So for some time I shared Zanottis interpretation. I will try to summarize it in order to be able to discuss it. I still share Zanottis characterization of Machlup ideas, but I think now, that they can fit more into the Semantic Conceptions frame than in Lakatos Methodogy of Scientific Research Programs. Zanotti points two key ideas in Machlups position: the notion of indirect testing and his conceptions of the fundamental assumptions in Economics which he derives from A. Schtz ideal types theory.

4 5

Although this concern with practicing economists is an important part of Machlups legacy. Zanotti (1991). Gabriel has become a dear friend, since then.

As Zanotti rightly states Machlups position was refined through his discussion with Hutchison, and his new understanding of Mises apriorism. This new understanding gave an alternative to Murray Rothbards well known position.6 Machlup's position For Machlup, the main problem in Hutchisons position was its naivety of not recognizing different levels of hypothesis. Machlup, instead, distinguished between them a group he called fundamental assumptions which are not testable but indirectly and in group, not in isolation. Therefore to an assumed change, there must be added the assumptions about the conditions which are empirical... They can be of three different kinds or types: A, B. and C. Assumed Conditions type A or type to case refers to conditions that are normally present. In Machlup`s words: these refer to conditions that may vary from case to case and influence the outcome significantly, but are sufficiently common to justify the construction of types (Machlup 1978 p.150) The examples include: types of goods involved, cost conditions, elasticity of supply or demand The Type B Assumed Conditions or type of setting in Machlup words "refer to conditions which may change over brief periods of time say, with a change of government or of the political situation, or during the business cycle- and are apt to influence the outcome in definite directions." (Machlup 1978 p.150) Examples of them are bank credit availability, fiscal and monetary policies, trade union policies, etc. Asssumed Conditions of the C type or type of economy, refer to "conditions which may vary from country to country and over larger periods of time, but may be assumed to be settled for a sufficient large number of cases to justify taking these conditions as constant" (Machlup 1978 p.151) Examples of them are social institutions, private property, social customs and usages, etc. Then, with the interaction of an assumed change, with the fundamental assumptions and assumed conditions of type A, B and C, a deduced change may be inferred. Only in the case of this deduced change being repeatedly disconfirmed by experience will fundamental assumptions be considered, through this indirect testing disconfirmed. Later, Machlup states clearly that the only flaw he finds in Friedman's classic article7 is that Friedman forgets that the fundamental assumptions of Economics "should be required to be understandable, in the sense in which man can understand the actions of fellowmen" (Machlup 1978 p.153).
6

Rothbard (1957) Rothbards article was written to state his differences with Machlup . In the article he calls Machlup several times and in different ways, a positivist. So in fact, the first of Misess interpretation was chronologically speaking, Machlups not Rothbards, however Rothbards understanding of apriorism has become standard. 7 Friedman (1953)

In one of his last published materials, (Machlup 1980) Machlup says about the discussion if models are a priori You may call any model a priori because you can build the model according to your own specifications Construction is always a priorieven if you construe with some experience in mind. The domain of construction needs constructs and postulated relationships between constructs, but it is itself not the result of observation; it is a prioriBut it is only a skeleton without flesh (Machlup 1980 p.1) Zanotti's interpretation of Machlup In support of his interpretation of Machlup as a Lakatosian precursor, Zanotti quotes an article in which Machlup even adopts the same expression than Lakatos, when he refers to fundamental assumptions as the "hard core" of a theory. And as the article is from 1974 Zanotti considers that Machlup has explicitly accepted his past position as very similar to Lakato's formulation of Scientific Research Programmes8. I do not deny that Zanotti seems to have a point, particularly if ones considers Machlup's semantic meticulosity of which his book on semantics (Machlup1963) is enough evidence, and his knowledge of Philosophy of Science. I even add to Zanotti's argument another astonishing linguistic coincidence; in his famous article on verification which sets the foundations of his position he states that other name that may be given to fundamental assumptions is "heuristic principles"9. Of course, this is not Lakatos' position which clearly distinguishes between hard core and heuristic instructions but proofs that Machlup was not far from it. Langlois' and Koppl 's position In their article (Langlois and Koppl) (1991) state the importance of understanding these more recent developments referring specially to Lakatos Methodology of Scientifi c Research Programs, among others. Understanding these more recent developments in philosophy of science is useful for understanding Machlups methodological views. For, in many respects, his position is precisely an anticipation of these modern developments10 Langlois and Koppl suggest that Machlups compromise between ultraempirism and apriorism was similar to Lakatos compromise between Popper and Kuhn. I agree that there is a resemblance but that is superficial: Machlup was a practicing economist with a serious knowledge of philosophy of science, trying to help stating some methodological guidance for other practicing economists. Lakatos was a philosopher trying to find a middle way between to other philosophers position, his main purpose was that. Machlups purpose was different he tried to make better economics. The authors note that Machlup uses sometimes model instead of theory but they dont stress the importance of that change, and the reason he gives for that change 11, by doing this I think they miss a very interesting point.
8 9

Machlup uses the expression while referring to the concept of "model". Machlup 1978, p.145 10 Langlois and Koppl (1991) p.88 Emphasis in the original. 11 See Section IV.

Langlois and Koppl analyze Alfred Schtzs influence on Machlup and conclude that his position is basically derived from Schutzs. I do not deny his influence, but I disagree in its degree. Nevertheless I subscribe to one of the authors conclusions: For one thing, Machlup recognized more clearly than Schutz, that our choices are not between individual models, but among theoretical systems -among hierarchical or quasihierarchical structures of inter-related models (Langlois and Koppl p.92) This is precisely a compelling reason to consider Machlup a precursor of the Semantic Conception, which can be rightly described as considering theoretical systems as hierarchical structures of inter-related models. Other opinions Wenceslao Gonzlez in his article (Gonzlez 2009) makes a brief reference to Machlup while analyzing Lakatos: quoting Spiro Latsis approvingly he states that Machlup's position is closer to Lakatos' idea of hard core, but he underlines also, that there is nothing in Machlup about predicton of new facts. In spite of these arguments, I believe that Machlup's position is better understood within the framework of the Semantic Conception in Philosophy of Science as a precursor of this family of thought in Philosophy of Economics more than a Lakatosian precursor. Before arguing in that way, I will briefly characterize the Semantic Conception in Philosophy of Science, for the reader who is not familiar with it. Therefore, anyone with a working knowledge of Philosophy of Science can safely skip next section and go straight to the fourth III. The Semantic Conception For the purposes of this work I will follow a quite standard textbook in Philosophy of Science from two authors that happen to be part of the Structuralist current, one of the main ones within the Semantic Conception. (Diez y Moulines 1999) They refer to the family of the semanticists as including some positions that, although their family features allow calling them part of a certain family, also show individuating differences. 12 I will concentrate particularly on those common features that allow including a particular position within the Semantic Conception, as I intend to show that Machlups work is better understood as a precursor of this particular family in Philosophy of Science. Starting from its chosen name, the Semantic Conception tries to establish its differences with the Received View that emphasizes the syntactic aspects of science. For the semantic conception a scientific theory is formulated by a set of models, not formulated by presenting a series of axioms, because theories are not only a set of propositions13. As a consequence of this enlargement of the concept of theory, a term can be labeled theoretical or non-observable only in reference to a theory T, therefore calling them ttheoretical. Thus it proposed a solution to one of the more difficult questions of the
12 13

The Wittgensteinian reference is, of course, intended by the authors. I will use proposition as used in modern logic of first order, also called propositional logic.

Received View, the definition of terms that were defined only negatively as nonobservables. And as the conception of model is fundamentally semantic, the chosen name intends to underline that idea.14 A classic work of the Structuralist branch of the Semantic Conception (Balzer, et.al 1987) says: The fundamental intuition underlying our approach is that the smallest significant or interesting parts of empirical science things like empirical laws- are best characterized, not as linguistic entities, but as model-theoretic entities classes of set-theoretic structures (Balzer, et.al p.xxi) Of course is a matter of emphasis, because it is not true to say that the Received View has overlooked totally the semantic aspects involved in Philosophy of Science 15 That is why the Semantic Conception must be understood as an intent to answer some critiques targeted originally to the Received View. The common family features in the Semantic Conceptions can be summarized in three characteristics: a) a theory is a set of models b) that set of models is identified by the empirical phenomena that it intends to explain c) that pretension, creates the empirical assertion of the theory, which consists in asserting that, between the models determined by laws and the empirical phenomena, there is some kind of relationship, that must be cleared. The first common feature states that theories are not a set of propositions, but a set of models. These models are determined by a set of principles, also called laws or usually axioms that point out to what entities or class of entities it is going to make reference. They delimit the universe of discourse. The second feature adds the empirical aspect, pointing the empirical phenomena that the theory pretends to explain. Models are not determined as a self-contained activity, but to explain empirical phenomena. These empirical targets are called intented applications(Balzer et.al) or intended scope (Suppe 1977) The last feature makes the theorys empirical pretension explicit; that there is a relationship between its models and empirical phenomena. This relationship can be of very different kinds according to different members of the Semantic Conception: can be one of identity, of approximation or of subsumption. But the point is that models fit the systems the theory wants to explain. In order to end this brief characterization of the Semantic Conception, another key consequence of considering a theory a set of models instead of a set of propositions is that truth can not longer be predicated about theories, that means that theories can not longer be directly true or false, because they are not propositions and truth or falsehood can be only predicated about propositions. The attribution of truth or falsehood can only be made to a theory indirectly thorough its empirical assertions, which are propositions.

14 15

Other usual way to refer to the Semantic Conception is as the Model-Theoretical Conception Particularly Rudolf Carnap wrote extensively about semantically aspects of the language of science in Carnap (1942) Carnap (1945),Carnap [1947] (1970) and of course in Carnap [1963] (1997)

IV. Machlup as a precursor of the Semantic Conception The key article in which the similarities between Machlups position and the Semantic Conception are clearly stated is originally from 1960. (Machlup [1960](1978). In that article in a part titled Construct and Model, Machlup states two points I will analyze with some detail: the definition and the form and composition of models. a) The Definition of Model The first point that gets Machlups attention is the definition of a model. Initially, he condemns its reckless and sometimes sloppy usage. Of course, none of these accusations can be directed to the Semantic Conception which shares with Machlup a careful attention to detail and precision. Machlup then provides his own definition of a model as a system of interrelated constructs and distinguishes it from a theory, stating that a theory is more than a model. His definition of theory is strikingly similar to the Semantic Conception. In his words:
A theory may be regarded as a model plus a specification of the empirical observations to which it applies16

And later, in the same page, he adds

Thus, to repeat applicable or applied theory consists of two parts 1) a pure model and 2) a specification of the empirical facts (described ordinarily in terms of real types or operational concepts) whose changes it will explain or predict.17

Machlups definition of theory is nearly identical to the Semantic Conceptions, it includes a model and an empirical aspect. There is not just a terminological coincidence, there are not just using the same terms, they are saying something very similar. b) The form and composition of models After suggesting taxonomy of models, Machlup states that the only pertinent for his analysis are the verbal and the symbolic ones, which he divides between geometric and algebraic. But states that in any model its function is to show connections, relationships and interdependences. In his words: But in all forms the function of a model is to exhibit connections, relationships, interdependences. There would be no reason for making a model except to show how some things hang together or of what elements they are composed or how they work or are adjusted18

16 17

Machlup [1960](1978) p. 178 emphasis added. Machlup [1960](1978) p.178 18 Machlup [1960]1978 p.176

Again, we can see similar characteristics with the Semantic Conception, since both state the key aspect of empirical phenomena in model building. As a difference I must point that some authors within the Semantic Conception (Moulines particularly) have repeatedly affirmed also the unavoidable intertheorethical aspect of model building. But this is not a feature that can be attributed to all members of the Semantic family. Of course these references, although very clear, are not of much value is they are isolated. Therefore, I will try to show next that this is not the case with Machlup. For example during his famous discussion with Hutchison (Machlup 1956) on referring to Hutchisons problem to understand the notion of indirect testing, he states that the root of the problem is Hutchinsons not acceptance of the existence of a third kind of proposition19 which although truth or falsehood can not be predicated of them, they still have empirical content. In his words: If so, he rejects a third category of propositions used in most theoretical systems: the heuristic postulates and idealized assumptions of abstract models of interdependent constructs useful in the explanation and prediction of observable phenomena. Such propositions are neither true or false nor empirically meaningless. They cannot be false because what they predicate is predicated about ideal constructs, not about things or events of reality. Yet they are not empirically meaningless because they are supposed to apply or correspond broadly to experienced events. They cannot be falsified by observed facts, or even be proved inapplicable because auxiliary assumptions can be brought in to establish correspondence with almost any kind of facts; but they can be superseded by other propositions which are in better agreement with these facts without recourse to so many auxiliary assumptions 20 Then in a foot note he points out that these were call procedural rules by Kaufman, complex analytical propositions by Montague and constitutive, non epistemic by Margenau.21 I think that this particular passage which I quoted extensively shows the main problem with sintacticism that reduces a scientific theory that is a complex social construct to a group of propositions. May be thats why Machlup as many others started later to talk about models, although he did not like much this polysemic term.22 In a previous work, Machlup after stating that models are indispensable for thinking although like the famous Monsieur Jordan of Moliere with prose, we may not be aware of using them, affirmed that there is no need to use that particular term model.
19

I called these linguistic entities propositions after Machlup, but by definition an atomic proposition is the smallest unit of meaning of which truth or falsehood may be predicated. 20 (Machlup [1956] (1978)p.496 21 Caldwell (1994) correctly states that this third category is unnecessary if the principle of indirect verification is applied. 22 For Machlup it was a weasel word and catch-all word ,in other words, a term that through its careless and polisemic usage, keeps it outside intact but is has no meaning inside, as an egg which was sucked by a weasel, See also Machlup (1963) p73-96 where structure is analyzed About model see Machlup (1978) p.175

Again, in his words: We think with the aid of models, whether we know it or not Of course there is not need to use the word model. Other words have been used in the same sense for example construct and schema23 Then in a later work (Machlup1961) Machlup quotes Richard Braithwaite and in his distinction between higher level hypothesis which can be never be directly nor indirectly verified isolately but together and the low level generalizations that can and are permanently tested in social sciences. Machlups quote is this: Only a whole system24 of hypotheses can be tested by deducing from some set of general postulates and some sets of specific assumptions the logical consequences and comparing these with records of observations regarded as the approximate empirical counterparts of the specific assumptions and specific consequences (Machlup,1978 p.354) Finally in one of his last published works on the matter, (Machlup 1978) Machlup makes an erudite description of different authors and their methodological positions. When he analyzes Hans Reichenbachs work he chooses to quote the preface of Experience and prediction and precisely Reichenbachs reference to indirect propositions which only their relationship to direct ones (referred to physical facts immediately observable) gives them some degree or probability. Its seems clear that Machlup saw the problems that some aspects of the Received View presented, and he tried persistently to solve them, basically creating new categories of propositions. The real limitation was the syntactic view that considered theories as propositions.

Conclusions Its difficult to deny Zanottis position that makes Machlup a Lakatossian precursor particularly because I once shared that view. Yet I believe now that Machlups similarity with the Semantic Conception is deeper. Of course this not means saying that Machlup was a crypto semanticist that will be an exaggeration of presents25, just to state that the well developed frame of the Semantic Conception in Philosophy of Science allows to continue, to expand and to complete some answers Fritz Machlups work aimed to.

23 24

(Machlup,1978 p.76) Machlups emphasis 25 Although a degree of presentism is unavoidable in every reconstruction, making an historical reconstruction impossible, its exaggeration is usually labeled as whiggish after Herbert Butterfields (1931) book.

These conclusions do no intend to be than a scholastic reinterpretation of Machlups texts but an approach trying to show that the Semantic Conception is compatible to and even anticipated by the Austrian Economists work. The main problems that Machlup confronted in the methodological area were originated in the bounds of the Received View. Those problems are clear in the necessity of introducing a third category of propositions which although having empirical content, truth or falsehood can not be predicated of them. Machlup followed Kaufman and Reichenbach on that, showing again his erudition in Philosophy of Science. The Semantic Conception solves the problem and removes the bounds of the Received View by changing its conception of what scientific theories are, making them complex social constructs26 and not just propositions. Machlup shared this point as I showed previously with a very important difference: Machlup was a part of an individualistic tradition and therefore always thought of models as individualistic mental constructs. The Semantic Conception denies that individualistic character and stresses instead the social aspect of this constructs. Although Machlup worked intensively in the production and distribution of knowledge27 - probably because of his individualistic formation- he missed the unavoidable communality of the scientific enterprise. It seems probable that Machlups rare erudition in the field of Philosophy of Science, made him aware of the shortcomings that the Received View was showing under the heavy criticism it was experiencing. He dealt also with the problem of the aprioristic position of his teacher, Ludwig Von Mises, which he cleverly solved by stating that models as mental constructs can be calle d a priori. It was clear that although Machlup was against using the term model for its polysemous character he stills used it to show that it was a different entity than a theory, because a theory was in Machlups definition a model plus a specification of the empirical observations to which it applies. This inclusion of the empirical intends is the Semantic Conception`s proposed solution for the famous theoretical terms problem This strikingly coincident definition of one the most important elements in Philosophy of Science the meaning of theory - is the main reason to reexamine Machlups works as an earlier precursor of the Semantic Conception. To do so, may help to establish a new understanding of Machlups works and, more importantly, to a new elucidation of todays practicing economists practice.

26

Semanticists still share Machlups search for clarity and content. Their conception of theories as complex social constructs does not give up on this aims. 27 Machala (1973)