The aim of this monograph is to provide a minimal metaphysics for scientific practice. A metaphysics for scientific practice is the project of making explicit assumptions concerning the structure of reality that best explain the success of scientific practice. A minimal metaphysics for scientific practice is a metaphysics that refrains from postulating any structure that is explanatorily irrelevant for understanding scientific practice.

The main argument of the book is developed by closely analysing paradigmatic aspects of scientific practice, in particular by focusing on the questions of whether and (if so) what metaphysical presuppositions best account for these practices. More specifically, in this book, I will look at the role scientific generalisations (laws of nature) play in predicting, testing and explaining the behaviour of systems, I will analyse causal reasoning, and I will examine reductive explanatory practices. These analyses yield the following results:

1. (1) An account of lawhood that takes law statements as attributing dispositional properties to systems. It differs from existing accounts by (i) taking the dispositions to be determinable, multi-track properties with a complex, functional structure that have laws of composition built into them, (ii) by its insistence on systems as the bearers of these properties – in contrast to authors who deny the existence of things or systems and in contrast to authors who assume that properties are the bearers of dispositions – and (iii) by spelling out natural necessity, including dispositional modality, in terms of invariance relations. Invariance is taken to be a modal notion that accounts for all other natural modalities encountered in scientific practice.

2. (2) A theory of causation in terms of quasi-inertial processes and interfering factors – concepts that can be explicated in terms of the account of lawhood developed earlier. Causal dependence can be understood in terms of nomological necessity and thus in terms of invariance relations. The account is a process theory that is able to cope with standard objections that have been raised against such theories: the problem of misconnection, the problem of disconnection and the reductionist assumption that all causation is ultimately physical.

3. (3) A theory of our reductive practices that makes minimal metaphysical assumptions. In particular, it rejects a foundationalist view that postulates a non-modal relation of ontological priority to account for an allegedly ontologically privileged level (e.g., as characterised by fundamental physical theories). I argue (i) that, contrary to appearances, such a view is explanatorily irrelevant for understanding our reductive practices and (ii) that arguments for a closely related eliminativist view that singles out physical facts as the only facts fails to be persuasive because they appeal to features of causation that are problematic on most accounts of causation. Scientific practice provides no reason to believe in Physical Foundationalism or Physical Eliminativism.

The rejection of an ontological hierarchy (as in (3)) suggests a new conception of what a science-informed metaphysics should look like. The positive picture that emerges is one that can be characterised in terms of ‘ontological monism’ and ‘descriptive pluralism’: It allows for a plurality of descriptions of a system (or of reality), none of which is privileged apart from pragmatic considerations. However, as descriptions of one and the same system, these different descriptions are constrained by a coherence requirement: We need to be able to understand how they fit together.

Metaphysics and Scientific Practice

My approach to the metaphysics of science distinguishes itself from other approaches in that it focuses on general features of scientific practice (such as prediction, explanation or manipulation) rather than on the content of physical theories. Many authors would agree that metaphysics (i.e., the study of the most general features of reality) needs to be informed by science. However, the details of this constraint remain controversial.

Science provides us with some of the best-confirmed claims about reality, and some of these claims seem to be directly relevant for metaphysics. So, there is a good prima facie reason to take science seriously as a source for metaphysics, as Maudlin notes:

There are, however, some problems with this idea. The first concerns the phrase ‘being informed by science’. It is sufficiently vague to allow for a number of models of the relation between science and metaphysics. This vagueness has been widely discussed (see, for instance, Reference HawleyHawley 2006, Reference PaulPaul 2012, Reference NeyNey 2012 and Reference LadymanLadyman 2012). I will not address this issue here (though I will take it up in Chapter 8).

Besides the vagueness of ‘being informed by science’, there is a further problem with taking the content of our best-confirmed theories as sources for our metaphysics; it is generated by a version of the argument from pessimistic meta-induction. Two of the candidates that should certainly inform metaphysical theorising – relativity theories and quantum theory – appear to be incompatible with each other. Theoretical physicists are still aiming at a unified theory. While it is surely true that a potential successor theory will reproduce the remarkable empirical successes of both theories, it is not evident that the ontological commitments of the successor theory will necessarily be very similar to those of the current theories. The history of science provides no evidence that such a condition needs to be satisfied. For instance, classical mechanics, which had a lot of empirical success that quantum mechanics was (and is) able to account for, provided no clue to the ontological lessons allegedly to be learned from its successor, quantum mechanics. Thus, given that quantum theory and the general theory of relativity are incompatible and in need of being superseded by a unified account, it appears to be fairly risky to rely on ontological lessons directly drawn from the content of these theories.

The point of these remarks is not that we shouldn’t trust scientific theories, or that we should completely refrain from drawing ontological conclusions from them. The point is, rather, that we need to be aware of the fallibility and limited reliability of this source in metaphysical investigations.

However, I would like to propose that besides the content of scientific theories there is another source that, while also fallible, might prove particularly fruitful for the metaphysical issues under consideration in this book. Moreover, I think that it may be a more reliable source than the content of scientific theories. The source I’m referring to is scientific practice. The scientific practice I have in mind is the practice of predicting, explaining, confirming and manipulating based on scientific findings. In what follows I will argue that certain aspects of this practice are best accounted for by making some very general assumptions about the structure of reality. In this sense, i.e., as the explanandum in an inference to the best explanation, scientific practice can be a source for metaphysical claims.

But why should this source be more reliable than the content of scientific theories? Certainly, scientific practice in general – just like scientific theories – changes dramatically over time: Scientists constantly develop new experimental methods for observing galaxies, for probing into tissues, etc. New methods or practices of representation have been devised in mathematical physics: delta-functions have been introduced, and renormalisations group approaches enabled new ways of coarse graining in statistical mechanics. True. But from a more abstract point of view some rather general features of scientific practice, such as part-whole explanations and extrapolation, remained unchanged even when quantum mechanics superseded classical mechanics. So, one thing I will assume is that there are certain features of scientific practice that are fairly stable over a long time and can therefore be taken to be a comparatively reliable basis from which to infer metaphysical conclusions.

Although scientific practice may not be a valuable source for all metaphysical issues, I will try to show in the following chapters that with respect to the issues I am interested in, i.e., lawhood, causation and fundamentality, scientific practice is indeed a reliable and fruitful source.

The project I have sketched is descriptive rather than revisionary, to use a distinction introduced by Strawson (Reference StrawsonStrawson 1959, 9ff). But it is not descriptive metaphysics in Strawson’s sense because it is not concerned with the structure of our thoughts about the world. The project is more ambitious in the sense that it is directed at what scientific practice reveals about the structure of the world itself. So, it is part of a descriptive, i.e., non-revisionary, but realistically conceived metaphysics.

But why should we assume that analysing scientific practice tells us something about the structure of reality whereas a prima facie analogous inquiry into the structure of our thoughts about the world gives us no more than very general features without which we cannot think about reality? The relevant difference is that – according to the Kantian or Strawsonian tradition – there is only one way we can think about reality or have objective experiences. By contrast, even though the features of scientific practice that I appeal to have proven to be fairly stable over time, there have been alternative practices of trying to cope with nature. These different practices had to prove themselves and did so with varying success.

In a letter Descartes wrote in 1638 to Morin, he appeals to a consideration of this type:

Descartes’ point is not only about rejecting certain ontological assumptions but also a claim about scientific practice: part-whole explanations are more successful than explanations that appeal to substantial forms, etc. Similarly, I take the very general features of scientific practice to which I will appeal to be part of a successful practice. By this I mean that these practices have proved successful with respect to predicting and manipulating the behaviour of the systems under investigation. Because of this success I take these practices to be good candidates for revealing something about the structure of the actual world. So, the main point is not to give a rationale or a psychological explanation of why we have a certain practice but, rather, to give the best explanation of its success, i.e., of why it works. It may, of course, be the case that the practices discussed here will be superseded by others. In this sense, the metaphysical claims advanced here are fallible.

Finally, let me mention that I am interested in scientific (rather than merely physical) practice. Even though most of my examples are taken from physics, the idea is that all kinds of disciplines may prima facie have a say in what is metaphysically presupposed about the world in their scientific practices. Quite a lot of science-oriented metaphysics looks for the most fundamental physical theories, in particular if it tries to infer metaphysics from the content of theories (Reference North, Albert and NeyNorth 2013, Reference NeyNey 2012, Reference FrenchFrench 2014, Reference Ladyman and RossLadyman and Ross 2007, Reference MaudlinMaudlin 2007). What is the rationale for this approach? Presumably something like the following argument: Metaphysics is the study of the most fundamental structure of reality. The sciences, too, are in the business of studying the structure of reality. Physics is the most fundamental of these. Thus, what physics has to say about the world is the best starting point for doing metaphysics (see, for example, Reference Maudlin, Loux and ZimmermanMaudlin 2003). However, as I will argue in more detail in Chapter 6, this view ultimately relies on an equivocation of the notion of fundamentality. Physics may very well be thought of as being fundamental in the sense of being, in principle, applicable to each and every system. Physics is the most fundamental discipline by being the most general. But it is a different question whether physics is fundamental in the sense that its subject matter is ontologically prior to the subject matter of every other discipline. Only when the latter assumption is taken for granted is it natural to assume that the theories (or practices) of fundamental physics provide the only clues for a metaphysics of science.

Overview

This book aims to give an analysis of some aspects of scientific practice and will in particular identify general assumptions about the structure of reality, which, I argue, best account for the success of our scientific practice. Most of the positive metaphysical claims are advanced in Chapters 1 and 2, while in Chapters 3 to 7 I argue that the structure outlined in Chapter 2 suffices to account for other aspects of scientific practice as well, e.g., our causal explanatory and our reductive practices. Almost no additional metaphysical structure needs to be postulated. In particular, the modal notion of an invariance relation, introduced in Chapter 1, will account for all the natural modalities and dependence relations we encounter in the scientific practice analysed in this book.

In Chapter 1, I start by analysing the role of generalisations in scientific practice. Law statements or generalisations are involved in one way or another in explanation, confirmation, manipulation or prediction. I argue that these practices require a particular reading of the generalisations involved, namely, as making claims about the behaviour of systems. These practices therefore presuppose the existence of systems or things.

Next, I look at the modal surface structure associated with laws. I use the term ‘surface structure’ to indicate that this structure may or may not be reduced to non-modal facts – as the Humean has it. I will sideline the debate about whether Humeanism is a tenable philosophical position. The positive claim I advance is that the modal surface structure can be explicated in terms of invariance relations – where I take invariance to be a modal notion.

In Chapter 2, I examine what appears to be a violation of the invariance of laws. Generalisations typically concern the behaviour of systems considered as isolated, while explanations, confirmations, manipulations and predictions typically concern non-isolated systems. Ceteris paribus clauses, which are often attached to law statements, take account of the fact that systems are typically not on their own. Systems are interacted on and interfered with by other systems – they are thus not invariant with respect to the behaviour of other systems. Understanding how ceteris paribus clauses work helps us to understand why we can explain, confirm or manipulate the behaviour of systems that are parts of a larger whole. An analysis of the role of ceteris paribus clauses shows that we need to read laws (generalisations) as attributing multi-track dispositional properties to systems. The argument relies on an analysis of scientific practice only and is not committed to more far-reaching claims, such as pan-dispositionalism, etc., that are common in the metaphysics of science literature. There is in particular no need to assume a sui generis conception of dispositional modality. On the contrary, I will argue that the modal aspects of dispositions can be explicated in terms of invariance relations.

In Chapters 3 and 4, I present a novel account of causation. On the one hand, it takes into consideration a certain plurality of causal concepts that has recently been diagnosed. On the other hand, it has two important features that are relevant for the overall argument of the book: (i) it shows how causation fits into a world as described by the scientific laws or generalisations, and (ii) it provides no reason to assume that causal relations are confined to a basic or fundamental level of reality – on the contrary, it might very well be the case that causation is a purely macroscopic phenomenon.

More particularly, I argue that our ordinary concept of causation (‘disruptive causation’) can be spelled out in terms of quasi-inertial processes and interferences. These processes and factors, in turn, can be fully explicated in terms of the generalisations provided by physics, biology or other sciences. The quasi-inertial processes in particular can be characterised in terms of the behaviour systems are disposed to in the absence of interfering factors and, thus, in terms of ceteris paribus generalisations and their underlying dispositions as discussed in Chapters 1 and 2. The analysis of the disruptive concept of causation is important for three reasons. First, this is the concept that is central for an understanding of why causal terminology plays an important role in scientific as well as everyday contexts. Second, this concept is tracked by most of our causal intuitions. I will argue that it fares much better than competing causal theories in accounting for these intuitions (this is what Chapter 4 is mainly concerned with). Third, I argue that the ordinary concept of causation is a focal concept: Other concepts of causation, which have been developed largely as a side effect of certain developments in the sciences and which do not explicitly rely on quasi-inertial processes and interfering factors, can be explicated with reference to this focal concept (in this sense, my explication takes account of a certain plurality of causal concepts). One example that will be discussed is what I call ‘closed system causation’: The state of a closed system at one time causing the system at a later time to be in a certain state.

As a result of this account of causation there is no need to make additional metaphysical assumptions beyond those introduced in Chapters 1 and 2. Causal relations can be fully accounted for in terms of generalisations such as those provided by physics, biology or other sciences; in particular, causal dependence can be explicated in terms of invariance relations.

In Chapters 5 to 7, I will turn to reductive explanatory practices. In the first part of Chapter 5 I distinguish various different ways in which reductive practices are important in some sciences. I will focus in particular on various kinds of theory reduction and of explanatory reduction. In the second part of the chapter I argue that the rationale for both theory reduction and explanatory reduction can be explicated in terms of our interest in achieving an understanding of how different descriptions of either one system (part-whole explanations) or classes of systems (theory reduction) are related. If we are confronted with different accounts of a class of systems (in the case of theory reduction) or of one system (e.g., accounts of the behaviour of one system in terms of different disciplines/vocabularies) we want these accounts to fit together. In the case of theory reduction we want to understand why a superseded theory was successful in the past and why we should continue to apply it. In the case of reductive part-whole explanations we want to understand why the description of the compound system and those of the parts and their interactions yield the same predictions ‘micro-macro coherence’).

Because we consider such an understanding to be an epistemic virtue, we seek theory-reduction and reductive explanations. I will discuss a number of cases (quantum entanglement, phase transitions, etc.) where we appear to have given up on reduction and coherence and argue that this impression is misleading.

The essential point in the context of a minimal metaphysics of scientific practice is that understanding the rationale for reductive reasoning requires an appeal to an epistemic virtue only and does not have to rely on metaphysical assumptions about, e.g., fundamentality or ontological priority.

In Chapters 6 and 7, I address whether there is a different argument for additional metaphysical structure. A quest for, e.g., micro-macro coherence is, surely, compatible with further metaphysical structure. Even if the rationale for part-whole explanations or other reductive practices can be spelled out without metaphysical assumptions and, this does not yet tell us why this practice is successful – why it works. To answer this question, I will look in more detail at Physical Foundationalism (Chapter 6) and Physical Eliminativism (Chapter 7) because it might be that, e.g., the particular way part-whole explanations work requires a metaphysical underpinning, which in turn provides evidence for one of these positions.

Chapter 6 will examine what exactly is presupposed about the metaphysical character of the relation between parts and compounds in the context of part-whole explanations. These kinds of explanations have often been taken to be evidence for Physical Foundationalism, a view that assumes that an ontological priority relation obtains between the micro-level and the macro-level. I will argue that part-whole explanations (just like other explanations) do indeed presuppose the existence of dependence relations between what the explanans refers to and what the explanandum refers to (this is sometimes called a ‘backing relation’). However, the stronger claim that an ontological priority relation obtains in nature does not do any work in understanding the dependence relations involved in our reductive explanatory practices. All we need is the assumption that parts and wholes mutually determine each other. A minimal metaphysics of science needs to postulate a dependence relation but not an ontological priority relation. Foundationalism is not implied by what classical mechanics and quantum mechanics have to say about the part-whole relation.

Nor is Physical Eliminativism – the view that the only facts there are, are fundamental physical facts – implied by our reductive practices, as I argue in Chapter 7. The positive picture that emerges is one that can be characterised in terms of ‘ontological monism’ and ‘descriptive pluralism’: It allows for a plurality of descriptions of a system, e.g., on a micro- and on a macro-level, none of which is ontologically privileged as the exclusively true account of reality, provided they are empirically adequate.

In Chapter 8, I will reflect on the methods and the character of the metaphysical arguments employed in Chapters 1 to 7. I will argue that most of the arguments can be reconstructed as inferences to the best explanation. Objections against the use of inference to the best explanation in metaphysics can be countered with respect to the specific contexts discussed in this book. What counts as the best explanation in these contexts can be understood in terms of what is minimally required for giving an account of the success of certain features of scientific practice. I will situate this approach by contrasting it with naturalistic and aprioristic conceptions of doing metaphysics of science.