7 hours ago
Review of Hempel on the "received view"
I haven't updated the blog for a while, but I have been reading Carnap, Hempel, Nagel and others on the syntactic conception of theories. It is quite striking how they appear caricatured by the next generation when you go back to the original texts, but doing so also sheds light on the semantic view, for we can see its premisses appearing in the later texts (in particular Hempel and Nagel). Here are a few comments on three articles from Hempel.
## Hempel (1952) "Fundamentals of concept formation in empirical science"
This is a long, but accessible article that presents very clearly the project of logical empiricism.
Scientific theories are able to unify a variety of empirical laws and suggest new ones (something presented vividly in Nagel's "The structure of science" (ch. 5) and Carnap's "Philosophical foundations of physics" (ch. 23)), but the way to do so requires introducing new technical terms, and the project is to _explicate_ these terms.
> In the initial stages of scientific inquiry, descriptions as well as generalizations are stated in the vocabulary of everyday language. The growth of a scientific discipline, however, always brings with it the development of a system of specialized, more or less abstract, concepts and of a corresponding technical terminology.
An example from another of Hempel's articles is how "wood floats on water" is less generalisable than "solids float on liquids of higher density", that uses a new technical term, density. We should _explicate_ it.
An explication is a concept from Carnap, and it corresponds to what we would call today conceptual engineering. Hempel explains:
> An explication of a given set of terms [...] aims at reducing the limitations, ambiguities and inconsistencies of their ordinary usage by propounding a reinterpretation intended to enhance the clarity and precision of their meanings as well as their ability to function in hypotheses and theories with explanatory and predictive force. (12)
From an empiricist point of view, this means connecting them to a vocabulary that is already understood. The article then goes through all the notorious difficulties of doing so, notably the fact that terms are often dispositional and the fact that theoretical quantities are defined on a continuum while observations are finite. In the end this forces us to be very liberal. We cannot expect correspondence rules to follow specific patterns, such as strict definitions or reduction sentences.
What is interesting with respect to the usual criticisms of logical empiricism is the following:
* Clearly, there is no question of first order logic; it's quite obvious that Hempel and Carnap (on whose work this article draw heavily) want to account for all advanced mathematics, including real numbers, etc.
* Both Carnap and Hempel consider the possibility of introducing modal operators ("causal modality") in order to analyse dispositions. Their only problem is, apparently, that modal logic was too poorly developed at the time. They are not radical Humeans!
* The basic observation terms are not purely psychological, they must be publicly accessible; they can go beyond direct observations to the extent that they are antecedently understood, which includes those used with high intersubjective agreement by scientific experimenters in measurements. (Carnap also claims that the division of vocabulary is partly conventional, and Nagel that it is a question of degree, see references above).
* Hempel and Carnap are of course fully aware that correspondence rules cannot take the form of strict definitions, but even if correspondence rules are liberalised, it is still possible to have well defined criteria of empirical significance to exclude pure metaphysics, so it is not a reductio ad absurdum of the project on their view.
* Actually, the liberalisation of correspondence rules is not viewed as a defect, it is actually viewed as an interesting development, for it can capture interesting features of science, notably the plasticity and indispensability of theoretical systems.
This last point is much more salient in Hempel's "theoretician's dilemma" article to which I now turn.
## Hempel (1958) The theoretician's dilemma
This is a more technical article focused on a specific problem. If a theory has empirical significance, this is because it has empirical consequences, which implies, paradoxically, that theoretical terms are dispensable: we might do as well with all the empirical consequences instead of the theoretical system. On the other hand, if a theory has no empirical significance, its terms are dispensable as well, because the theory is useless. So, why do we need technical terms in science?
After a long detour through many different issues (correspondence rules, Craig theorem, etc.) Hempel's response lies precisely in the fact that correspondence rules are not strict definitions or reduction sentences. This means that theoretical terms cannot be substituted by equivalent empirical formula. This is what provides their value: they are _plastic_ , the theory--world interface can be extended or revised in many ways, guided by its postulates.
Here I was quite surprised by how realist, quite close to externalists like Kripke or Putnam indeed, he sounds in the following passage:
> whatever observational criteria of application the scientist may provide are intended by him to describe just symptoms or indications of the presence of the entity in question, but not to give an exhaustive characterization of it. The scientist does indeed wish to leave open the possibility of adding to his theory further statements involving his theoretical terms--statements which may yield new interpretative connections between theoretical and observational terms; and yet he will regard these as additional assumptions about the same hypothetical entities to which the theoretical terms referred before the expansion. This way of looking at theoretical terms appears to have definite heuristic value. It stimulates the invention and use of powerfully explanatory concepts for which only some links with experience can be indicated at the time, but which are fruitful in suggesting further lines of research that may lead to additional connections with the data of direct observation.
Now the crucial drawback of this liberalisation is that correspondence rules cannot be taken to be _meaning_ postulates, since they have empirical consequences! For example, if we extend a theory to new measurement techniques, this assumes that old and new techniques give the same result when both are applicable, but this is an empirical claim. This was already noted in the 1952 article, as well as before that by Carnap.
If correspondence rules are not meaning postulates, but empirical claim, not analytic, but synthetic, we have a problem of intelligibility for theoretical terms. They are no more defined. A possibility, considered by Ramsey and later Carnap, is to bite the bullet. Theoretical terms are _not_ intelligible, and therefore, we should not use them! At least for reconstruction purposes, we should quantify on predicates instead. This is the original role of Ramsey sentences.
But Hempel isn't satisfied, and leaves this problem open at the end of the article.
## The rise of the semantic view (Hempel "On the 'Standard conception' of theories" , 1970)
Twelve years later, Hempel returns to the problem of intelligibility of theoretical terms, and there are interesting newcomers in the discussion. He mentions notably Nagel, for whom a theory is associated not only with postulates and correspondence rules, but also with a _model_. In Nagel (The structure of science, 1961), this mostly corresponds to the mental pictures that we associated with the theory, for example, we imagine the molecules of a gas behaving like billard balls (there are antecedents of this idea in Hesse's "Models in Physics" article from 1953, cited by Nagel) . The point is, even if the connection to direct experience is thin, there might be other ways to make theoretical terms intelligible. He cites Sellars:
> the conceptual texture of theoretical terms in scientific use is far richer and more finely grained than the texture generated by the explicitly listed postulates
Another related aspect is that Hempel is more critical of the division of vocabulary between observational and theoretical, noting that terms such as "position" are both observational and theoretical. Again, this could help in "picturing" the theory.
Now comes the fatal move to go to a semantic view. What if theoretical terms were intelligible only in virtue of the theoretical postulates, and associated model, without any need for an empirical interpretation? What if scientific theories were not empirical systems, but purely structural ones? Theoretical laws mere implicit definitions of these terms, in the same way as the axioms of artihmetic are sometimes taken to define numbers implicitly, as whatever satisfied the axioms?
This move is attributed by Hempel to Suppes. In effect, this is a reversal of the intuitive idea that theoretical postulates (e.g. the laws of Newton) are _synthetic_ (they claim something about the world) while correspondence rules linking them to experience are analytic (they define them). We saw that the latter must be false. Then perhaps the postulates are actually analytic, because to be a Newtonian force is just something formal, to have some conceptual relations to Newtonian mass, which is just as formal. These are nodes in set-theoretical models. The synthetic claim is actually provided by correspondence rules, to the effect that some part of nature is also a model of the theoretical system, or isomorphic to one.
We see this kind of framing in many semanticists (to the extent that some of them even deny that a theory can be true or false: it can only be applicable or not!). This is what is really meant when semanticists claim that we should characterise a theory by means of their models: we should take the theoretical postulates to be implicit _definitions_ of abstract structures, analytic statements that is, and correspondence rules or whatever postulate of theory--world correspondence to carry all the cognitive content.
Although I already got the main idea behind the semantic view that postulates describe structures, I found it illuminating to be able to situate it in this way in relation to the problem of intelligibility of theoretical terms that affects the syntactic view. We can see that in a sense, externalism about meaning, neo-essentialism and causal theories of reference are in direct concurrence with the semantic view (since they provide a different solution to the same problem): something that is not often noted! This made the semantic view more palatable to me, and I guess I should be thankful to Hempel for this.
I migrated and united all my blogs to a new address, and also posted a brief review of three Hempel articles on the "received view" of theories.
personal.us.es/qruyant/blog/2026/04/rev...
#philosophyofscience
0
1
0
0
