Review of "There Are No Such Things As Theories", by Steven French

In “There Are No Such Things As Theories”, Steven French defends that there are no such things as theories. This well-named book provides a careful review of a wide range of issues, ranging from the philosophy of science (the syntactic and semantic conceptions of scientific theories, fictionalism about scientific representation, theoretical equivalence, science in practice) to the philosophy of art, with a particular focus on the metaphysics of abstract objects, including fictions and works of art, and, of course, theories, so as to arrive at its provocative conclusion, which is roughly the following: scientific theories don’t really exist, but we can still make sense of ordinary discourse about scientific theories.

I wasn’t convinced by this conclusion in the end, but I think that this is because I approach things very differently at the meta-philosophical level, and I am not particularly moved by metaphysical considerations in general (the ones signalled by the “really” emphasised in the sentence above). Still, this was a good read (particularly chapter 8: “theories in history and practice” for my purpose).

Summary of the book

The two first chapters review the debate between the syntactic view of theories and the semantic view. As a reminder, the syntactic view roughly considers scientific theories to be statements about the world, while the semantic view considers them to be families of mathematical structures that can be compared to the world (a subtle distinction that, according to some, ends up being superficial).

In chapter 1 (Theories as Sets of Propositions), French helpfully distinguishes, following Hendry and Psillos (2007), a weak and a strong version of the syntactic view: only the latter insists on a partial interpretation of theoretical statements via correspondence rules linking a theoretical and an an observational vocabulary, as per the logical empiricist approach championed by Hempel and Carnap, while the former is happy to interpret theoretical statements “literally” (whatever that means!). This is important, because one of the main motivations for a semantic view has been problems with correspondence rules, in particular the fact that they would imply that theories change every time experimental techniques evolve, which seems implausible. As explained in chapter 1, focusing on a weak version avoids this difficulty, and this is still the implicit conception of theories in many areas of philosophy. French is thus careful to avoid easy caricatures, noting for instance that a syntactic view is not in principle incompatible with the idea that models play a role in science (this is rather clear in Hendry and Psillos’s article, which I recommend to anyone interested in the debate).

Chapter 1 then mentions two further possible understanding of the syntactic view, in terms of sentences or in terms of propositions, and argues that there are issues with each options, in particular metaphysical issues in the case of propositions (What are they? If they are abstract, how do we have access to them?). Finally, it examines another difficulty for the syntactic view, which has to do with the fact that it is committed to theories directly representing the world rather than being compared to it: how to account for idealisations, and the way models mediate between ideal theories and messy phenomena in practice?

Chapter 2 (Theories as Families of Models) is about the semantic conception, of which the author is an advocate, but with an important twist: French acknowledges that “anything the Semantic Approach can do, its Syntactic rival can do also”. This has to do with recent attacks against the semantic view, notably by Hans Halvorson (2012) and Sebastian Lutz (2015), on which I might comment later on this blog. The former in particular claims to show that the way the semantic view identifies theories, by means of their structure alone, is wrong: it sees equivalence where there is none and conversely. According to French, the arguments are question-begging because the inadequacy is judged on the basis of syntactic criteria, and also non-conclusive because they are applied to toy theories: this is forgetting that theories are ultimately representational entities, something semanticists, despite their emphasis on structure, never denied. He ultimately thinks, as far as I understand, that both the syntactic and semantic approaches are valid: this is not a substantial decision, but a practical one, because none of these views are about what theories really are. They are mere representations of theories for some analytic purpose, they are not ontological theories of theories. Yet for practical purpose, according to French, the semantic view is superior in many respects.

Chapter 3 (Theories as Representations) is about scientific representation. There, French presents the partial structure approach that he has been advocating in many other places as well (very roughly the idea that some elements of models are faithful, others not and others we just don’t know, all this explained in structural terms), and he explains how it can account for idealisations and other features of representation. There is a philosophical debate about what makes something a representation of something else (my own views are sinthesised in Ruyant 2025), with many comparisons between art and science, and notorious arguments to the effect that representation cannot be just a matter of similarity or morphism: too many things are similar to many other things, so some intentional components must be involved to fix the directionality of representation. French objects first that art and science are different (in art, the author of a representation is more important for example, whereas in science, it only has historical significance), and second that the intentional component isn’t very clear in the case of science. A realist and a positivist would differ about what a theory represents for example. All this to defend the partial structure approach. And again, exhibiting the formal structure of representation is not the same as taking an ontological stance, so this still doesn’t tell us what theories really are.

Chapter 4 (Theories as Abstract Entities) presents the main challenge that is discussed in the rest of the book, which concerns the idea that theories would be abstract entities: how can abstract entities be discovered or created? The challenge applies as well to musical work: is Beethoven’s fifth symphony created or discovered? The former we would say, but abstract objects cannot be created. Or can they? Here is the challenge in the context of theories: at least one of these propositions must be denied.

• Scientific theories (and models) are abstract objects.
• Scientific theories (and models) are created.
• Abstract objects cannot be created.

The subsequent chapters address this challenge. Chapter 5 (Theories as Abstract Artefacts) in particular discusses whether theories, conceived of as abstract entities, could be created, using Popper’s three worlds theory and Amie Thomasson’s theory of abstract artefacts: objects that lack spatio-temporal location, but are still created, change and can cease to exist, all this on the basis of our intentions. French is not convinced by Thomasson’s theory. One of his main concerns is that scientific theories (or pieces of music) are developed gradually. There is a fundamental problem of identification: what exactly is the theory, when does it come to existence? Another difficulty that is more visible in the case of science is that they can be independently discovered by different scientists, at least it seems, which doesn’t fit well with Thomasson’s focus on authors’ intentions. All this is discussed carefully and at length, both in the case of art and science.

The alternative is to think that theories are discovered instead of created: we approach them little by little during the development process, sometimes many scientists do it at the same time, but they existed all along in an abstract realm. But here, the worry is that truth apparently constitutes an important constraint on theories, so much that, according to him, it’s awkward to say that a false theory was “discovered”. But true and false theories all exist on a par in the abstract realm.

If all options that retain the premise that theories are abstract are unsatisfying, then maybe theories are not abstract entities. Chapter 6 (Theories as Fictions) addresses fictionalism, and in particular a recent trend that consists in applying Walton’s theory of fiction to scientific models: models would be “props in a game of make-believe”, prescribing our imagination of physical systems by means of rules of generation. This is an ontologically minimalist solution that doesn’t require theories to exist in an abstract realm, but according to French, it seems implausible due to the fact our imagination is limited in a way theories and models are not (we can use computers to discover their content for instance).

At this point, the problem remains unsolved. Chapter 7 (Theories Eliminated!) finally offers the author’s preferred solution: eliminativism. The solution is a direct transposition of the same solution to the same challenge in the case of musical work, proposed by Ross Cameron (2008). It is based on a distinction between ordinary language and ontologese, or between ordinary statements and statements that describe the world at its more fundamental level. “There are theories” might be true in ordinary language, because it has truthmakers at the fundamental level, but these truthmakers might not be theories and so the same statement is false in ontologese, or shouldn’t be interpreted as expressing ontological commitments. Theories do not really exist, but what makes ordinary talk of theories true is a set of practices. As far as I understand, the main point is that these truth-makers could be variable from one context to another, so that ordinary language need not be consistent, and so, we can preserve the three propositions of our trilemma.

Chapter 8 (Theories in History and Practice) provides further arguments for the view by examining issues of theoretical equivalence and theoretical developments. Science is messy: there doesn’t seem to be strict criteria of identity for theories, or precise points in time where we can say that the theory is there. But of course, this is not a problem if theories do not exist. Then chapter 9 (Theories in the Realism Debate) goes “meta” so to speak, by examining philosophical discourse about theories, in particular debates on scientific realism. According to French, the truth-maker theory would allow to preserve scientific realism even if theories do not exist, because the truth-maker of philosophical truth talk are philosophical discursive practices.

A few comments

This book helped me figure out what I’m not interested in. I’m interested in what theories are, but not in an ontological sense. I would like to have precise answers about when we can say that two theories are equivalent or the same, that one reduces to another, that a theory is confirmed by experience, that it is true or acceptable; I would like to know what is their main functions for their users, how they relate to models and how they should be interpreted in general, and so on. And for sure, some of these issues are addressed in the book. But in my view, once all these questions are answered, there is nothing more to ask. If we have all these answers and you tell me “OK, fine, but we still don’t know what a theory really is”, I’ll just give you an incredulous stare. What else do you want? But that’s exactly what the book asks after reviewing some of these issues: what theories are, really, at the ontological level. And the conclusion, that they do not exist, is an answer to this precise question. This is why although I enjoyed many of the discussions along the way, I often felt the urge to give incredulous stares while reading and couldn’t buy most of the final conclusions of the chapters. Only when I reached chapter 8 did I feel that the real issues were addressed at last, and that a real challenge was mounted against the idea that there are theories. But the solution provided didn't convince me. However, I think this is really a question of meta-philosophical stance.

One reason why I don’t feel the urge to address “deep” ontological issues is that the same kind of question can be asked about representations in general, about propositions, linguistic meanings, beliefs, thoughts, pieces of art, or about mathematical objects, or about norms, values, etc. So, it seems to me that the answers won’t be specific to scientific theories. This is why I think it is legitimate to just understand theories in terms of these other philosophical concepts without saying more. If it makes sense to say that theories are propositions for instance (and I think it does), then, to me, the inquiry stops there: “proposition” is a widely used term of art with rather clear connections to other terms of art or more ordinary concepts such as truth and meaning, which makes it legitimate for us, philosophers of science, to use as well, and we can delegate deep ontological questions to philosophers of language, of mind, or other philosophers interested in general metaphysics. When we reach the deepest levels, nothing specific to science remains, it seems to me, and so, there is nothing to worry about in the context of the question “what theories are?”. Doubts about propositions are not legitimate, because if the concept is so pervasively used, there must be a way of making sense of it. We might discuss which way fits more with scientific discourse, but questioning the idea that theories are propositions on the ground that they are mysterious isn't fair. The same goes when it comes to abstract objects, norms, or any general philosophical notion we would like to invoke to understand scientific theories. In other words, I would have liked some of the discussions to be more specifically about scientific theories (but to be fair, some were).

But I guess that even when it comes to general metaphysics, I personally tend to be sceptical about this kind of ontological inquiries. The very idea of Ontologese, for example, doesn’t make much sense to me. All languages are on the same boat, and this particular language, Ontologese, has never been heard or spoken. In the end, it seems that it only amounts to adding “really” before any existential claim, whatever the language we’re using, and we might wonder what exactly is meant by this addition. My answer would be: nothing.

To say it differently, in my view, saying that theories don’t exist, as the book does, is tantamount to saying that we should stop talking in terms of theories. It is saying that we should find better concepts if we want to understand science and analyse the activity of scientists. We should even forget about the semantic view and the syntactic view, then. This is how I implicitly understood the title before reading the book. But this is not at all what is claimed: in the end, philosophers and scientists can continue to talk about theories as usual, and to analyse them using one or the other view. We can add “really” to convey the specificity of the claim that is made, in contrast with my initial understanding, and then we see that it only affects how we should talk when speaking Ontologese, not ordinary English: as for the latter, business as usual. But the book is written in English, and nobody speaks Ontologese, so the claim becomes quite irrelevant to me. (Now I’m fully aware that the whole idea of speaking Ontologese is somehow metaphorical for expressing ontological commitments, but this does nothing to dispel all these worries; you can read all my qualms in a metaphorical way if you want).

I guess the problem is that we don’t have the same way of fleshing out what ontological commitments are: to me, they can be mundane (expressed in English or whatever human language) rather than deep. We don't need to "go meta" and ask what we "really" mean when we speak.

There is indeed a tendency in various places of the book to go meta, and meta again, and again, that I typically want to resist (and which is far from specific to this author of course: I feel the same when reading the work of Kit Fine for example). So, for example, it is claimed about the semantic and syntactic views that “the framework merely represents, at best, what we call theories and models; what they are, ontologically, should not be read off the framework, lest we end up regarding theories as set-theoretic entities” (p. 178). Well, it seems to me that the semantic view and the syntactic view describe what theories are, that it is what representing something means, and I see no difference between describing an entity and making ontological claims about it. Let's keep it simple and first-order. If the semantic view claims that theories are collections of set-theoretical structures (the kind of thing set theory describes), then so be it: this is an ontological claim about theories. Why not interpret these views literally? If there are no theories, but only dis-unified practices, which seems to be the conclusion of the book, then perhaps we should update our representations of science, and replace the semantic view with something else.

Regarding this tendency of "going meta", I found the last chapter, where it is suggested, roughly, that philosophers engaged in the debate on scientific realism speak English and not Ontologese, even harder to swallow. If really the distinction between Ontologese and other languages made sense (which, of course, I doubt, but if it did), then surely, philosophers engaged in the realism debate would want to talk Ontologese, not English, and particularly realist philosophers! Otherwise they are not really realists. It seems to me that eliminativism about theories isn’t compatible with realism simply because the realist wants a robust notion of truth, not a reinterpretation of it as is proposed. So, scientific theories must make substantial claims about reality, and then, this is what they are: something propositional presumably, however fleshed out in metaphysics. But, you will say, the claims aren't as well defined as it seems? Fair point (perhaps), but at this point, an error theory, which would actually constitute an original version of anti-realism (theories can’t be true because they don’t exist), becomes much more plausible than the way of saving realism that is defended (roughly that philosophical discourse shouldn't be interpreted literally). Philosophical discourse is already "meta" in a sense: it is a discourse about theories, which are discourses about nature. We should be able to speak literally about theories at this level (just as some urge us to interpret theories literally when they talk about nature). Going meta again about it is just too much for me.

Finally, I didn’t really understand why the idea that the truth-makers of theory-talk would be elements of practice don’t suffer from the same problems discussed at length in the book: what is practice? When is it scientific? Are the relevant elements type-level or instance-level? Actual or potential? Are they normative, intentional? What is a norm? etc. The account doesn’t shed much light on theory-talk in the end. Saying that there are no theories, but truth-makers for theory-talk, sounds like putting the problem of what theories are under the rug.

All this to say that I have quite a lot of disagreements with the content of the book, but many of them come from my general meta-philosophical approach. In the end, the book contains many intersting discussions for philosophers interested in understanding representation in science, fictionalism, conceptions of theories, etc. or for philosophers who, unlike me, are interested in deep ontological questions.