Jump to ContentJump to Main Navigation
PhysicalismThe Philosophical Foundations$

Jeffrey Poland

Print publication date: 1994

Print ISBN-13: 9780198249801

Published to Oxford Scholarship Online: October 2011

DOI: 10.1093/acprof:oso/9780198249801.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 24 October 2019

The Theses of Physicalism

The Theses of Physicalism

(p.186) 4 The Theses of Physicalism

Jeffrey Poland

Oxford University Press

Abstract and Keywords

This chapter formulates a set of theses that ultimately aim to satisfy the criteria of assessment set up in Chapter 1. Thus, they must be theses that are sufficient for expressing the core ideas of the physicalist programme while avoiding the pitfalls encountered by other formulations. They must be capable of withstanding the numerous objections launched against the programme by its opponents. And they must fare well in the face of relevant evidence. Consistent with the negative assessment of the formulations considered in Chapter 2, the theses formulated are responsive to physicalist concerns about emergence and explanation.

Keywords:   physicalism, physicalist doctrine, realization, emergence, explanation

All right-thinking physicists believe that macro-thermodynamical quantities, states and events are parasitic on the microscopic. But it turns out to be hard to characterize the parasitism in terms of identities (e.g. ‘temperature is mean kinetic energy’ is a glib over-simplification of a very complicated relation which may not in the end turn out to be a relation of identity).

John Earman (1986: 249–50)

A reductive explanation not only allows one to deduce the laws of the secondary theory, but also to tell us what it is about the nature of the objects of the secondary theory that brings it about that they have the properties they have and what it is about their nature that brings it about that those properties are related to each other in the way the laws of the secondary theory assert them to be related.

Berent Enc (1976: 303)

The goal of this chapter is to outline a set of theses which gives adequate expression to the core physicalist ideas and values discussed in Chapter 1. The theses must suffice for expressing the physicalist's vision of one of the important ways that things hang together (viz., physically-based unity of the universe and of knowledge that concerns it). Thus the ideas of ontological dependence, supervenience, and realization must be captured in a way that reveals, in general, how all phenomena are embedded in the physical fabric of the universe, and that circumscribes the scope and limits of influence and interaction. Principles expressing the physicalist's belief that all objective facts and truths are determined by physical facts and truths are required. And the idea that all that happens is explainable and integratable within a system of physically-based vertical explanations must be framed. In these ways, the theses should concern metaphysical and epistemological structures that (p.187) will yield understanding and that can play an important role in the pursuit of research and in the development of a culture.

Two further sorts of consideration are important for pursuing this objective. First, we must heed the morals of Chapter 2 regarding the shortcomings of other formulations. The issue of how properly to express concerns about realization and non-emergence, on the one hand, and vertical explanation, on the other, must be confronted and resolved, given the demise of classical reductionism and given the essential weakness of the various supervenience relations. Second, as will become evident, there are a number of objections to the theses that must be dealt with if the programme of physicalism based upon those theses is to get off the ground. The objective is to frame theses that jointly provide an adequate expression of physicalist ideas and values, that do not fail for reasons comparable to those that undermine other formulations of physicalism, and that are not vulnerable to further objection.

4.1. Ontological Dependence, Supervenience, and Realization

A formulation of the physicalist ontological position must give expression to the idea that the physical ontology is basic with respect to all other phenomena. I have suggested that being basic in this context has three components: ontological dependence, supervenience, and realization. My arguments in Chapter 2 were aimed at establishing that not only do the various versions of classical reductionism fail to express these ideas, but also a host of supervenience theses fail as well. Reliance upon purely linguistic formulations, purely supervenience formulations, and theses involving nomological sufficiency or equivalence is not enough for the physicalist's ontological purposes. More powerful alternatives must, therefore, be introduced.

Before proceeding with the development of such theses some preliminary comments are in order. As noted earlier, it is necessary to treat all the various ontological categories in a full development of the physicalist position. However, I shall not do this in the present project for reasons of space. Rather, the focus will be on objects and attributes. There are a number of points worth noting with respect to each of these categories. (p.188)

To begin with, there are serious problems in how to treat objects within a physicalist system: the viability of the concept is brought under deep suspicion by developments in contemporary physics. However, I shall bypass all issues concerned with, for example, criteria of individuation, location, and boundaries of objects. Whether and how such problems, at both the level of microphysics and more macroscopic levels, are to be resolved must be left open, but they must be acknowledged as a debt of the physicalist programme. Further, I shall subsume under the concept of an object all manner of concrete individuals, whether they are natural kinds, well-defined artificial kinds, or rather unnatural objects such as the fusion of some arbitrary class of objects. Such distinctions are important for the working out of the physicalist programme, as different categories may well be treated differently, but for present purposes a more abstract approach suffices. As I shall discuss in more detail shortly, the physicalist ontological position can be adequately expressed in terms of regions of space-time and their contents, where the latter are construed in terms of the instantiations of objects1 and attributes.

I also make a number of assumptions about attributes. My focus will be on qualitative, as opposed to non-qualitative, attributes: i.e. I shall focus on attributes that do not involve any essential reference to particular individuals. And my use of the term ‘attribute’ should be clearly understood to include both properties and relations. Within the former category, I include both intrinsic and extrinsic properties. As argued in Chapter 2, I see no insurmountable difficulties with the idea that an individual may possess a certain property in virtue of how things are elsewhere in the world (for example, the property of being the tallest person).

Along a different line, I take it as obvious that there are non-physical, as well as physical, attributes. Therefore I reject as false the view that every attribute is a physical attribute. This rejection is, I believe, the received wisdom resulting from the struggle with these issues over the past few decades in the philosophy of mind and elsewhere. Rather than leading to the demise of the physicalist programme, however, such wisdom should lead us to appreciate more deeply the ways in which the physical is basic. Further, I (p.189) reject the view that, while denying the identity of all attributes with physical attributes, asserts token identity of attributes. This view founders either because it leads to the view that there are property particulars or that somehow it is possible for properties to be distinct but their instantiations to be identical: neither option is viable, in my opinion.2

Finally, recall that the development of the physical basis for ontology in Chapter 3 led to characterizations of (a) classes of physical objects and attributes, and (b) the class of possible physical states of the world each consisting of space-time regions and their physical contents (i.e. distributions of physical objects and attributes).3 And similarly to the case of the physical domain, other ontological domains can be understood as being constituted by classes of objects and attributes, the members of which can be distributed over regions of space-time. Within each domain, classes of basic objects and attributes are identified, and classes of complex and higher-order objects and attributes are generated from the basic classes via appropriate generative operations.4

The distinction between the actual distribution over regions of space-time of the objects and attributes in an ontological domain and the range of possible alternative distributions in that domain can be understood in a way comparable to how that distinction was made for the physical domain. However, the characterization of this range of alternatives is more problematic than for the physical case since, in addition to constraints imposed by laws appropriate to that domain, there are other constraints as well. For example, ‘laws’ concerning non-physical attributes need not be without exceptions. As a consequence, in order to clarify the range of possible alternative distributions of objects and attributes in a given non-physical domain, it is necessary to clarify the constraints upon such exceptions. I shall return to this issue below.

Given this framework of assumptions concerning ontological matters, the formulation of the first ontological thesis of physicalism is as follows: (p.190)

  1. (T0) All objects and attributes that are (or can be) instantiated must be instantiated in regions of space-time.

This is the traditional physicalist demand that if any object or attribute, basic or complex, physical or non-physical, is instantiated in this world, then it must be instantiated in the ‘space-time causal nexus’. Alternatively put, everything that is real in nature occurs in space-time. In addition, T0 constrains the possibility of an object's or an attribute's being instantiated: all that is possible in this world includes only those phenomena that can occur in space-time.5

The thesis presupposes (and I shall discuss this at length in Chapters 5 and 6) that the natural order can be distinguished from the realm of abstracta. And it relies upon the assumption that space-time physics is a relatively enduring feature of our conception of nature. Should that conception change dramatically, then some other way of pinning down the natural order will have to be introduced. Finally, note that T0 does not rule out ghosts and their ilk, if they can be located in regions of space-time. Although T0 provides an important ingredient of the physicalist's ontological view, it is clearly not the whole story.

The second physicalist ontological thesis is roughly that, associated with the actual (or possible) instantiation of any entity (i.e. object or attribute) in some region of space-time, there is a class of physical entities that are also actually (or possibly) instantiated. As stated, however, this is too rough. Focusing on attributes first, the mere co-occurrence of some class of physical attributes with a given non-physical attribute does not provide much bite to the idea that the physical ontology is basic. I have not even stipulated (nor will I) that the physical attributes and the given attribute are instantiated in the same region.6 So, on the assumption that there is always some physical attribute or other which is instantiated, we have the consequence that, unless further conditions are added, (p.191) any attribute which is instantiated in a region of space-time will trivially satisfy this second thesis in virtue of satisfying the first thesis. What more, then, is required of the association between physical and non-physical attributes to establish the former as basic?

It is here that the physicalist must call directly for the realization of a given non-physical attribute, N, by the members of some class, P, of physical attributes with which it is associated. Nothing less will do, as I argued in Chapter 2. Thus the second ontological thesis is as follows:

  1. (T1a) For each non-physical attribute, N, and for each region of space-time, R, if N is actually (or possibly) instantiated in R, then there exists a minimal class of physically-based attributes, P, such that the instantiation of the members of P does (or would) provide a realization of N on that occasion.

A realization of an attribute, N, on a particular occasion by a class of physical attributes, P, is a configuration of physical attributes that constitutes N on that occasion. How the members of P accomplish this will vary depending upon the nature of N. For example, the realization of the property of transparency in terms of a given physical structure that permits light rays to pass through it will differ significantly from the realization of the content of a mental state in terms of (say) causal relations between a symbol and the objects or attributes to which the symbol refers. By focusing on particular instantiations of the attributes involved, T1a allows that consideration of time and place can be relevant to the realization of a non-physical attribute (for instance, when an object's having a certain history is relevant to the constitution of one of its attributes), Clearly, it allows for there being different classes of physical attributes each of which provides a realization for some given N on different occasions: T1a is compatible with the multiple realizability of attributes.

In calling for physically-based attributes as members of P, T1a requires that all realizations of non-physical attributes are ultimately grounded in instantiations of physical attributes. But it accommodates the idea that there are many levels of ontological complexity and abstraction and that, at the higher end of the range, realizations of certain attributes consist of non-physical attributes. For example, (p.192) the instantiation of properties of human behaviour will in many cases involve properties of bodily motion plus properties of a larger physical or social environment (as in the case of a bodily motion's being the signing of a cheque). What T1a requires is that all properties and relations involved in the realization of a certain non-physical attribute must be themselves either physical attributes or physically-based attributes. The ontological picture is one of a hierarchy of attributes that is grounded in the physical basis and that is structured by the relation of realization. Many intermediate levels of attributes may be required for the instantiation of some high-level non-physical attribute.

T1a also stipulates that the class P must be a minimal class of physically-based attributes. The idea is that there can be no free riders: each of the members of P must make an essential contribution to the realization of N. Thus I take it to be a burden of the form of physicalism I advocate that a distinction can be drawn between a class of relevant attributes that combine to constitute a realization of N, background-sustaining conditions that, although not constituting N9 are required for its instantiation on a given occasion, and totally irrelevant attributes that neither constitute N nor provide significant background. For example, there would be no people on Earth if Earth were 20,000,000 miles from the sun. But that relational property is not constitutive of the attribute of being a person, although it is a background-sustaining condition of there being any persons on Earth. On the other hand, a certain very tiny rock on Mars having a mass of 2 grams is quite irrelevant to the instantiation of personhood on Earth (i.e. it is neither constitutive of personhood nor is it a significant background condition for the instantiation of personhood).

Although the class of physically-based attributes called for by T1a should include only those attributes that play a crucial role in constituting the non-physical attribute N on a given occasion, there is no general way in which the required distinctions can be drawn. What is relevant to the constitution of an attribute, what is significant background, and what is irrelevant are determined on a case-by-case basis. And it is quite compatible with T1a that there are holistically realized attributes (i.e. attributes which are realized by total states of the universe). These points serve to indicate that physicalism is compatible with a wide range of ways in which higher-order phenomena can be realized. However, the (p.193) class, P, will in general consist of two subclasses: those attributes which are directly relevant to the constitution of N on the given occasion of instantiation, and those attributes which make up the background-sustaining conditions of N on that occasion. Both classes are essential to the instantiation of N, but in different ways. In the case of holistically realized attributes, of course, all physically-based attributes that are instantiated will be included in the first subclass.

It should be noted that realization is an essentially asymmetric relation and therefore it has at least one of the marks of a genuine dependency relation. Of course, it has more than that since realization is a relation which directly captures what is central to the dependency of the non-physical upon the physical: the latter constitutes the former and provides the conditions necessary for the instantiation of any attribute in this world. Thus the problem of relevance that plagued CR and the various forms of non-reductive physicalism is handled effectively by T1a. We can, for example, distinguish between a class P that realizes an attribute N and a class of nomic equivalents of the members of P that are each irrelevant to the instantiation of N: the former class has members that constitute N, whereas the latter class does not. The problem of emergent properties is also readily dealt with: realization is the direct antithesis of emergence. In requiring that the former relation holds, the latter relation is ipso facto precluded. Neither nomological sufficiency nor any species of supervenience relation can make that claim.

On the other hand, the realization of an attribute also provides, inter alia, a nomological sufficient condition for that attribute (i.e. a connective generalization which connects the joint instantiation of the members of P with the instantiation of N is true).7 Although I argued in Chapter 2 that this is not enough for the physicalist, I want to side with those who include nomological sufficiency as among the relations required to obtain between non-physical attributes and attributes in the physical base. The realization (p.194) relation must have at least this much modal force if it is to avoid what is surely not in the spirit of the physicalist programme: viz., that two relevantly similar physical states of affairs should constitute realizations of incompatible non-physical attributes.

In T1a, the parenthetical qualification concerning ‘possible’ instantiations of an attribute expresses the idea that physicalist ontological theses bear upon both the actual and the possible states of this world. T1a asserts that any attribute that either is or can be instantiatiated in this world must satisfy the stated condition. In other words, as I shall argue in Chapter 5, T1a is not a thesis about other possible worlds. Rather, it is only the possible states of this world that fall within its scope. Any attribute that could be instantiated in this world and hence is an aspect of some possible state of the world must be associated with a class of physical attributes the members of which are also aspects of that possible state and are jointly sufficient for realizing the target attribute.

T1a is a reductive thesis in so far as it requires that there be associated with the instantiation of any non-physical attribute, N, a specific class of physical attributes that are co-instantiated with N. Such reductionism, though modest, is indeed a form of reductionism.8 It requires association of non-physical attributes with specific classes of physical attributes, subject to the condition that the latter provide realizations of the former. It therefore differs significantly from non-reductive supervenience claims of the sort discussed in Chapter 2. Further, since T1a requires no strict correlations or identifications of non-physical with physical attributes it is not a version of either CR or CR'. None the less, the condition constraining the association of non-physical attributes with classes of physical attributes is designed to give expression to the idea that the physical ontology is basic within nature. However, as I shall discuss below, this constraint is thought to be vulnerable to the objections that it is too weak to determine unique associations and that it is too strong to be plausible. Thus there are many who object to even this modest form of reductionism. (p.195)

Regarding the relations between T1a and various supervenience theses, T1a entails the global supervenience claim that if A and B are possible states of the world, alike in all physical respects, then they are alike in all non-physical respects. This is because A and B, ex hypothesi, will agree on all the physical attributes relevant to the realization of any non-physical attribute that is instantiated in those states of the world. Thus T1a captures the idea that non-physical attributes supervene upon the physical attributes in this world. Again, although supervenience of this sort is not sufficient for expressing key physicalist ideas, it is an essential constituent of any adequate formulation of physicalist doctrine.

Further, T1a implies the weak supervenience thesis that if two individuals agree on all their physical attributes (intrinsic and extrinsic), then they will also agree on their non-physical attributes. Again, this is because the individuals will agree on all relevant physical attributes bearing upon the realization of whatever non-physical attributes are instantiated. On the other hand, T1a does not entail at least one version of the local supervenience of the non-physical on the physical: i.e. it does not entail that any two regions of space-time that are alike in all intrinsic physical respects must be alike in all non-physical respects. This is for the reason that many non-physical attributes are realized by relational properties that extend beyond the region in which the given non-physical attribute is instantiated.

It should also be kept in mind that T1a associates the non-physical attribute N with a class of physically-based attributes, not with a single attribute. Thus, it is distinct from what Kim calls ‘strong supervenience’. T1a does indeed imply strong supervenience if complex physical properties can always be constructed from the attributes in P, and if I am right that there is a relation of nomological sufficiency between the instantiation of those attributes and the instantiation of N. However, strong supervenience clearly does not imply T1a since the relation of nomological sufficiency is much weaker than realization, as I argued in Chapter 2.

Finally, whereas T1a calls for the least class of instantiated physical attributes that provides a realization of each instantiated non-physical attribute, Post's local supervenience thesis, LS1, calls for the least class of physical attributes that suffices to determine a given non-physical attribute. T1a implies, but is not implied by, LS1, again because, whereas the truth of T1a guarantees the (p.196) existence of those conditions that make LS1 true, realization is a much stronger relation than determination.

With respect to objects, I suggest the following physicalist thesis:

  1. (T1b) For each non-physical individual object, O, and for each region of space-time, R, if O is actually (or possibly) instantiated in R, then there exists a minimal class of physically-based objects, PO, such that the instantiation of the members of PO does (or would) provide a realization of O on that occasion.

The thesis, which focuses exclusively upon individual objects (as opposed to kinds), allows for the existence of non-physical objects. But it none the less imposes a significant constraint on what individuals can be instantiated in the world.

To say that some objects are ‘non-physical’ means that such individuals are instances of kinds of object that are not among those identified in physics or generable from such kinds (i.e. they are not among those in the physical basis for ontology as outlined earlier). Rather, they are instances of kinds constituted by, and individuated in terms of, various non-physical attributes. For example, tables and chairs, although realized by physical objects in our world, are non-physical in character. They are not found in the physical basis and they are constituted by various non-physical attributes (for instance, functional attributes). Further, social institutions, cultural artifacts, and works of art are all additional examples of non-physical objects that are realized by classes of physical objects in one type of organization or another.

Such non-physical objects can be natural objects (for instance, biological cells) or artifacts (for instance, pencils). They can be relatively simple or complex: for example, such biological kinds of objects as single cells and complex living systems are both non-physical objects that must be located within the physicalist framework. Furthermore, such objects can be highly unnatural fusions of objects of the same or different kinds: the fusion of any class of physical or non-physical objects is itself an individual which must satisfy the demands of T1b. Thus the conception of reality that the physicalist system must accommodate is one which countenances, inter alia, the many kinds of natural and artifactual objects that we encounter in daily life and that we identify in (p.197) various theoretical and other cultural endeavours. The point of endorsing T1b is to identify the physical conditions that any individual must satisfy to be real. But in doing this, the physicalist is not ruling out all the many highly diverse aspects of our experience or the many kinds of objects that populate our world.

There are, of course, problems with the boundaries of objects (cf. Quine 1981b) and hence there are problems with their location in space-time regions. A resolution of such problems needs to be developed, either in terms of a stipulation that resolves the lack of clarity or in terms of tolerance of ‘fuzzy’ objects. I do not see that there are any serious problems for the physicalist programme with regard to either of these strategies of resolution. Given T1b, the coordination of classes of lower-level physically-based objects with individual, higher-level non-physical objects is the principal problem here that must be solved as the physicalist programme is worked out. This is indeed a difficult problem, but its difficulty does not arise primarily because of problems with boundaries.

As with the condition on attributes stated in T1a, T1b provides a condition for both actual and possible instances of non-physical objects. In no possible state of the actual world can there be an individual object that is not realized by some class of physically-based objects. From works of art, social institutions, and linguistic utterances to molecules, cells, and living creatures, each individual of these kinds of object is subject to the demands of T1b. On the other hand, ghosts, gods, and demons, as generally conceived, do not and cannot exist in this world if T1b is true because none of these sorts of beings are realized by elements grounded in the physical domain.

Again, as with T1a, T1b calls for associating a given non-physical object, O, with some minimal class, PO, of physically-based objects. The minimality condition signals the commitment to drawing distinctions between those objects in the world that are directly relevant to the realization of O on a given occasion, those objects that are part of the background-sustaining conditions for O on that occasion, and those objects that are irrelevant to the realization of O. Although there is no general principle or strategy for drawing such distinctions and although there may be a certain amount of arbitrariness in how the boundaries of O are circumscribed, there do not appear to be any insurmountable problems for (p.198) physicalism posed by this commitment. Considerations of relevance and irrelevance are based upon the kind of object O is and the particular way in which O's boundaries are drawn.9 And again, as with the class P called for by T1a, PO will consist of two subclasses, the first involving objects that are directly constitutive of O and the second involving objects that are part of the sustaining background.

In calling for classes of physically-based objects, T1b allows that the realization of certain kinds of non-physical objects (for instance, social institutions) involves systems of other non-physical objects (for instance, persons). Given some non-physical object, O, it is some configuration of other non-physical objects (and their attributes) which constitutes the realization of O in this sort of case. And indeed, each of those other objects may in turn be realized by configurations of non-physical objects. The point of stipulating that it is physically-based objects which go into the realization of O is that all objects that realize O are themselves realized by physically-based objects and that at some point the chain of realization comes to an end in the physical basis.

As we have seen, the fundamental metaphysical relation that structures the world as understood by the physicalist is realization of an object O by a class of other objects PO. It is not identity of an object O with some physical object. There may, of course, be plausible identity statements relating an object in one theoretical domain with an object in another. But the stock philosophical examples of such identities (for example, water, genes) do not underwrite a general identity approach, even if they are plausible.10 Physicalists who are interested in encompassing all objects that occur in nature, not just a highly restricted class within the domains of the various natural sciences, must recognize that identification of, say, cultural objects like a work of art or an artifact is not plausibly identifiable with any physical object. Realization is a relation that is based upon the idea that particular objects exist in nature as a consequence of how certain types of more basic objects (and their attributes) are organized and embedded in a larger context. Again, it is the job of realization theories to clarify what constituent objects and what types of organization (p.199) and contextual embedding are required for realization of some particular kind of individual that happens to exist in some space-time region.

To sum up, T1b calls for an ontological structure in which all objects are embedded. The structure can have many levels of organization that are asymmetrically related by a relation of realization and that are all ultimately grounded in a basic level of physical objects. How objects of some given kind are realized can vary, but the critical stipulation is that, in all cases, an object exists in the world only if it is realized by a class of physically based objects. How a given object is in fact realized depends upon the details of the relevant realization theories as well as of the particular facts of the case. Thus the critical stipulation is a way of ruling out ghosts and their ilk while accommodating both the wide diversity of kinds of objects in the world and the implausibility of identifying all objects with physical objects. Along with T1a, T1b puts content and bite into the physicalist's metaphysical vision. These theses jointly entail that everything that occurs in nature is dependent upon, supervenient upon, and realized by what goes on in the physical domain.

4.2. Determination of Fact and Truth

The second class of physicalist theses are those concerned with objectivity. Based upon the discussion in Chapter 1, I make the following assumptions. Objectivity is ordinarily taken to be a property that attaches to ‘truth’, ‘knowledge’, and ‘facts’. That is, it is taken to be a feature of the world as well as a feature of our knowledge of the world, and physicalists are concerned to capture both employments of the concept. Further, there are two key ideas relating to the notion of objective truth and knowledge: on the one hand, inter-subjective testability and agreement and, on the other, independence of languages, theories, and minds. The issues here run quite deep, since there is an important sense in which all knowledge depends upon subjective features (for instance, representational systems). As a minimum, however, the notion of independence alluded to is intended to rule out obvious individual features upon which certain claims may depend (for example, specific personal goals that motivate claims to knowledge), and it (p.200) is intended to capture the idea that what is true is not subject to an individual's interests and desires.

For example, Quine's thesis of the indeterminacy of translation is a claim concerning the non-objective status of translation from one language to another (‘there is no fact of the matter’). Even if such translations are objective in the sense of being supported by evidence available inter-subjectively, they are not, according to Quine, objective in the sense just described: different translators with different interests, sensitivities, and proclivities could, as a consequence of those differences, translate differently but equally correctly. Such sensitivity to individual differences is a feature of translation but not of physics, according to Quine. The reason for this divergence requires a third idea regarding the notion of objectivity: the idea of there being an objective matter of fact underlying a claim to knowledge.

According to Quine, translations do not count as objective knowledge because they are sensitive to individual differences between translators: they are not independent of the knowing subject. Physics is not sensitive to such differences: it is independent of the knowing subject. What distinguishes those claims which exhibit such independence from those which do not? Quine's, and the physicalist's, answer is that the former concern matters of fact while the latter do not. Hence physics, being concerned with matters of fact, is not sensitive to differences among investigators, while translation, being not concerned with such matters, is sensitive to such differences. This is why two translators, who differ in subjective features, can, on the basis of those differences, offer different translations which are equally correct. There is no underlying fact of the matter that makes one translation right and the other wrong.11

It is in this way that the physicalist's interest in objectivity focuses on the idea that the independence of certain claims from subjective factors can be understood in terms of there being matters of fact with which ‘objective’ knowledge is concerned. Such matters of fact offer up critical resistance to subjective variation of interest, sensibility, and the like. The world is the way it is whether we like it or not and whether we know it or not. To capture these ideas (p.201) within the physicalist programme, two questions must be addressed: ‘Within a physicalist conception of nature, what counts as an objective matter of fact?’, and ‘Within a physicalist system of knowledge, what are the features of objective truth and falsehood?’ (i.e. ‘Within a physicalist system of knowledge, how are claims with a factual basis different from those that lack such a basis?’).

The appropriate physicalist strategy for answering these questions is to require certain relations to hold between all domains of fact and truth and the physical bases. In particular, the physicalist theses concerning objectivity will characterize relations which must hold between non-physical phenomena and physical phenomena in order for the former to count as objective matters of fact, and between non-physical claims and physical claims in order for the former to have an objective truth value. If both these sorts of thesis are formulable and defensible, they should constitute an expression of the physicalist view that the physical ontology and doctrine provide bases for objectivity.

With respect to matters of fact, the physicalist view is that the objective matters of fact are the physical facts and any facts that are determined by those physical facts. The problem of formulation thus comes down to the problem of making precise the physicalist dictum that the physical facts determine all the facts of nature. The core idea behind this dictum is that there is an independent world of fact that has a certain structure: i.e. it is structured by relations to a physical basis. Such relations constitute the conditions of objectivity within a physicalist world. If a purported fact can be shown not to exhibit such relations, then that is a sufficient ground for the physicalist to reject it as not being objective.

In formulating the physicalist view of objectivity, I have spoken of ‘facts’ and invoked the distinction between ‘facts’ and ‘non-facts’. Some readers may balk at such expressions, and a few words of clarification might be helpful. With respect to the use of ‘fact talk’, I endorse the view that such talk is completely dispensable in favour of other locutions. Below, I shall present a strategy for doing this. With respect to the distinction between fact and non-fact, I am sympathetic to the idea that the conception of any fact is dependent upon features of the conceiver. For example, the nature of representation and the differences between representational systems are quite pertinent to the nature of cognition, and (p.202) hence to the differences between various cognitions of the world. But this general dependency does not undermine the possibility of drawing distinctions between fact and non-fact in what is conceived. From within any way of conceiving of knowledge and the world, distinctions between objectivity and subjectivity can be clearly delineated, even if that way of conceiving, like all others, involves parochial features of cognitive and representational systems. Thus, within the physicalist view of nature and knowledge, the conception of objectivity described above can be articulated.

To return now to that conception: if the objective facts are those that are determined by the physical facts, how is this to be expressed in a precise way? This question raises three more: ‘How are we to characterize the physical facts?’, ‘How are we to characterize the non-physical facts?’, and ‘How are we to characterize the relation of determination that the former bears to the latter?’.

As outlined in Chapter 3, a possible physical state of nature is a complete12 distribution of physical objects and attributes over the regions of the space-time continuum compatible with the laws of physics. The class of all such distributions is the class of all possible physical states of nature. I shall represent this class with a class of models, each member of which is a space-time structure in which there is one of the possible, complete distributions of objects and attributes over space-time regions. This class represents all the possible distributions of physical fact.13

To represent the non-physical facts, the strategy is similar. Consider the class of all non-physical objects and attributes and the space-time continuum as before. Then, a possible non-physical state of nature is a complete distribution of the non-physical objects and attributes over the regions of the space-time continuum which is compatible with the laws of nature.14 The class of all such (p.203) distributions is the class of possible non-physical states of nature. As before, I shall represent these possibilities in terms of a class of models, each member of which is a space-time structure over which there is a complete distribution of non-physical objects and attributes. This class of models represents all the possible distributions of non-physical fact.

Given these two classes of models, it is possible to specify a third class of models, M, each member of which is a space-time structure over which there are complete distributions of both physical and non-physical objects and attributes, where the distributions are constrained by the laws of nature. This class represents all the possible pairings of the possible physical states of nature with the possible non-physical states of nature. In other words, this class represents all the possible states of nature. Given this class of models, we can characterize the relation of determination pertinent to clarifying the physicalist view that the physical facts determine all the facts of nature.

Informally, the relation of determination is expressed by the idea that once the physical facts are ‘fixed’, so are all the non-physical facts. The official statement of this idea is as follows:

  1. (T2) For any possible complete distribution of the physical facts, there is exactly one possible complete distribution of the non-physical facts.15

In terms of the class of models M, if A and B are members of M and if A and B agree on the distribution of the physical objects and attributes over the regions of space-time, then A and B must also agree on the distribution of non-physical objects and attributes over those regions. Thus according to T2, for any possible physical state of nature, there is associated with it exactly one possible non-physical state of nature. A more formal characterization of the relation of determination which structures this class of models will not be given here, although there do not appear to be any special difficulties involved in providing one (cf. Hellman and Thompson 1975; Post 1987).

To summarize: on the physicalist view, an objective matter of fact is one which is determined by physical facts. The possible (p.204) states of nature, which exhaust the possibilities regarding objective matters of fact, have been represented in terms of a class of models, M, each member of which pairs distributions of physical and non-physical objects and attributes. If M is structured in accordance with T2, then it is structured by a relation of determination (i.e. members A and B cannot agree on the physical facts while differing on the non-physical facts). Thus, for example, any alleged attribute for which it is contended that, in identical total physical states of the world, it might in one case be realized while in another not, is not an objectively real attribute of things. Such an attribute will not be among those included in the class of non-physical attributes employed in the model-theoretic construction described above. Rather, it will be rejected, on physicalist grounds, as unreal.16

I note that, although T2 is a consequence of T1,17 the two theses give different sorts of insight into physicalist thought. Whereas T2 formulates a global constraint upon total states of nature and upon objective matters of fact, T1 provides (among other things) crucial insight into why T2 holds. The often sought-for understanding of why global supervenience theses are true of the world is that physical objects and attributes constitute realizations of non-physical objects and attributes. In observing this, I am not supporting those who believe that theses like T1 are required in order that global supervenience relations can be explained. There are more fundamental reasons for requiring T1, as I argued in Chapters 1 and 2. As Post has very ably discussed, there is no particular reason why global supervenience (or T1, for that matter) requires explanation at all. To think otherwise may be to hold on to some version of the erroneous principle of sufficient reason.18 My claim is only that, as a matter of fact, T1 does provide insight into why T2 holds.

A second physicalist thesis concerning objectivity bears upon knowledge directly, rather than upon the objects of knowledge. The question to be addressed is: ‘What are the marks of objective truth?’ For the physicalist, this is to ask: ‘Which knowledge claims (p.205) have a basis in physical fact for their truth value?’ (i.e. ‘Which claims have a truth value because of the way the world is physically?’). And the other side of the question is: ‘How are such claims different from those which are not objectively true or false?’ Thus the physicalist seeks a condition for objective truth and falsity. One of the virtues of classical physicalist reductionism was that it provided such a condition: viz., derivability from the physical truths was considered a sufficient condition for being an objective truth, derivability of the negation of a sentence was deemed sufficient for the sentence's being an objective falsehood. With the demise of classical reductionism the physicalist must formulate a viable alternative.

In recent years, the principle that the physical truth determines all the truth about nature has been favoured by many physicalists. It has generally been agreed that classical physicalist reductionism is too strong a thesis and that, however the principle of truth determination is to be cashed in, it must be compatible with the falsehood of such reductionism. In Chapter 2, I discussed some of the proposals that have been suggested for meeting this demand. For present purposes, I shall only offer a relatively informal version of the determination thesis; more technical developments can be found elsewhere in the physicalist literature (Friedman 1975; Hellman and Thompson 1975; and Post 1987).

Assume that a class of languages, L, is fixed and that the class of models, M, described above provides interpretations for those languages. The thesis of truth determination is as follows:

  1. (T3) Within the class of models, M, if members m 1 and m 2 agree with respect to the physical truths, then they agree with respect to the truths, formulable in all other languages in L.

This version is similar to that offered by Hellman and Thompson (1975), although it is less technical and differs from their formulation in other ways.19

How does T3 serve the motivations of the physicalist programme regarding objectivity? A sentence, s, formulable in some language (p.206) Li, has a legitimate claim to objective status only if its truth value is fixed once the physical truths are fixed: i.e. for s to be objective, it cannot be possible that, given a specification of the physical truths, either s or its denial could be true. Thus, in the terms of T3, there could not be models in M which, while agreeing on the physical truths, do not agree on the truth value assigned to s. As a criterion of objectivity, T3 purports to capture the idea that it is the mind independent physical facts that determine the truth value of s. Thus a clear distinction between objective truths, on the one hand, and judgements sensitive to idiosyncratic subjective features, on the other, can be drawn.

In this formulation, ‘truths’ are viewed as sentential rather than as propositional and non-linguistic. Thus it presupposes that the physical language is sufficiently powerful for expressing the physical truths that determine all other more complex or higher-order physical and non-physical truths that are expressible in the languages of L. If this condition is not satisfied, then the thesis could very well be false, for the relatively uninteresting reason that a sufficient number of the physical truths (propositionally construed) are not expressible in the physical language that happens to have been selected. If no such language exists, then T3 is false for a more interesting reason, although not in a way that undermines central physicalist ideas: in particular, T2 could still be true. Thus it is only relative to this assumption regarding the physical language that T3 is true and that it is a consequence of T2.

4.3. Vertical Explanation and Realization Theories

Earlier in this chapter, I presented ontological theses requiring that there be associations between instantiations of non-physical objects and attributes and classes of instantiated physically-based objects and attributes which constitute their realizations. This strongly suggests that explanatory relations should exist between the physically-based phenomena and the non-physical phenomena they realize, and that the ontological theses characterize a structure which provides a framework for developing such explanations.

More fundamentally, in Chapter 1, I represented the goals of the physicalist programme as including the promotion of a number of explanatory objectives the attainment of which would increase (p.207) the level of understanding that a physicalist system of knowledge yields. At the core of these explanatory aims is the provision of vertical explanations of the realization of non-physical objects and attributes. Thus all phenomena are to have explanations in terms of physically-based phenomena. A system of knowledge structured by such vertical explanations will be a physically-based explanatory system in the sense that all explanations, of whatever sort, will only make appeals to objects and attributes that are ultimately grounded in the physical domain, via a chain of vertical explanations. Such grounding further provides a framework for the explanation of all interaction and influence, as well as a delineation of their scope and limits. And it provides a framework for the integration of explanatory structures drawn from different domains (i.e. all such structures are embedded within the physical domain). Thus, significant constraint relations between theories developed in various branches of knowledge are implied by this aspect of physicalist thought.

In the present section, I shall formulate two theses which will serve to express the physicalist's ideas and motivations concerning explanation and which will clarify the nature of the vertical explanations suggested by T1 and sought after by the physicalist. The morals drawn in Chapter 2 must be taken seriously here. In particular, exclusive reliance upon the apparatus of classical reductionism or the various forms of non-reductive physicalism must be avoided, since explanatory structure of the sort we are after is not provided by either definitions, derivations, supervenience relations, or relations of nomological sufficiency. In addition, the implausibility of both generalized identity theses and generalized eliminative theses means that an alternative approach must be pursued. Specifically, the theses must directly call for explanations of the required sort (i.e. explanations of how non-physical phenomena are realized by physically-based phenomena).

The physicalist's basic conviction is that all phenomena occur in nature in virtue of what goes on in the physical domain. Thus the physicalist believes that a thing has the attributes it has in virtue of what it is made of, how it is put together, and how it is related to other things. In other words, we can come to understand why a thing is the way it is and why it behaves in the way it does (for example, why it exhibits the regularities it does) in terms of these three sorts of considerations. Theses T1 and T2 serve to (p.208) capture this idea from an ontological point of view. The question remains how it is best expressed from the point of view of a system of knowledge structured by physicalist principles. My approach will be to frame theses that express the following idea: given that all individual phenomena, all regularities, and all instances of and exceptions to regularities which occur in nature occur in virtue of physical phenomena, there are physically-based explanations of all such phenomena.

By ‘individual phenomena’ I mean instantiations of objects and attributes in regions of space-time. By ‘regularities’ I mean nomological associations between instantiations of attributes in regions of space-time, where the associations may express causal or non-causal relations of covariation among attributes and where they may be deterministic or probabilistic in character. Mention of ‘exceptions to regularities’ reflects the idea that higher-level laws of nature are typically not exceptionless, although they are counter-factual supporting to some extent. This feature of certain laws need not, and ought not, be viewed as a flaw in our knowledge so much as a deeper reflection of the structure of things. Physicalism provides a natural framework for understanding why this is so.

The first explanatory thesis is as follows:

  1. (T4) All instantiations of non-physical objects and attributes are vertically explainable20 in terms of physical or physically-based objects and attributes.

By a ‘vertical explanation’ (VE) of an attribute21 is meant a certain sort of explanatory relation between attributes at different ontological levels, where such levels are distinguished by degrees of complexity or abstraction. Such vertical explanations can be seen as providing answers to questions of the following form:

  1. 1. In virtue of what lower-level attributes did the instantiation of such-and-such higher-level attribute occur?

  2. 2. How did such-and-such lower-level attributes constitute an instantiation of such-and-such higher-level attribute? (p.209)

Thus a VE is supposed to identify a class of relevant lower-level attributes that may be drawn from one or more lower levels, and to provide an account of how those attributes combine to constitute an instantiation of the target higher-level attribute.22

As T4 requires, the physicalist is concerned with there being such vertical explanations for all non-physical attributes. More specifically, the call for VEs is a call for ‘token’ explanation of the specific instantiations of all such attributes, in terms of lower-level attributes that are either physical or physically based (i.e. attributes that are themselves ultimately grounded in the physical basis by a chain of realization relations). Thus it is possible, and likely for many higher-level attributes, that none of the attributes cited in a vertical explanation are physical attributes; the requirement is only that they should be physically based. For a physicalist, then, a full answer to questions (1) and (2) with respect to the instantiation of a given non-physical attribute consists in citing certain relevant physical or physically-based attributes and in explaining that instantiation in terms of them. It should be noted that there is no requirement that different instances of the same attribute are vertically explainable in exactly the same way.

Such vertical explanation must be clearly distinguished from etiological explanations which identify the causes of some effect. Realization, constitution, and the like are not causal relations in any standard sense. Further, a VE is not in general a form of explanation in physics, even if physical attributes are involved. The principles that characterize how physically-based attributes combine to constitute an instantiation of an attribute are not necessarily physical laws, although such laws could, on occasion, be relevant to a VE. It is therefore a mistake to suggest that the physicalist aims at understanding all phenomena in purely physical terms, if that means that all phenomena can be comprehended within physical theory. This is a twofold mistake because VEs need only appeal to physically-based attributes and because such (p.210) explanation can involve non-physical principles that relate lower-level and higher-level attributes.23 The fact is that there are high-level, non-physical structures, processes, etc. which are instantiated in nature, and a physicalist system of VEs aims at showing how such phenomena can be accounted for in a physically-based world. It is not, however, the job of physicists to accomplish this objective.24

To characterize more fully the kind of explanation called for here, I shall introduce the notion of a realization theory (RT). Such an idea is implicit in much of the work in the sciences and in a variety of discussions in philosophy. The idea is that of a theory, associated with a given attribute, N, that abstractly characterizes the kinds of attributes that are sufficient for the realization of N and that shows how such attributes can combine to actually constitute N in specific cases.25 Examples of such theories include a characterization of transparency (cf. Chapter 1), a functionalist account of mental states, and a causal theory of reference. In all such cases, an appeal is made to a theory of how the instantiation of certain relevant physically-based attributes suffices for the instantiation of the target non-physical attribute.

For the physicalist, given an RT for an attribute, N, and given a distribution of physically-based attributes, vertical explanations can be generated to account for the specific instantiations of N, if they occur. Unlike the VEs that they help to construct, an RT for an attribute is a generic, abstract account of how an attribute can be constituted. Such an RT plays the crucial roles of helping to isolate and identify relevant physically-based attributes and of showing why those attributes are relevant and how they combine to constitute N. Thus RTs are essential for effectively answering questions (1) and (2) above. In this way, then, RTs are central to (p.211) the physicalist project of explaining how non-physical phenomena are possible in a physically-based world (i.e. the project of showing how the physically-based facts about the world constitute the non-physical facts).

For example, functionalist accounts of mental states delineate the pattern of causal relations which, if instantiated in some physical or non-physical system, suffice for the instantiation of a given mental state, M. The RT in such a case consists of a characterization of the required pattern of causal relations without appealing to any more specific kinds of relations that might exhibit such a pattern (i.e. it is neutral with regard to what processes, if any, underlie the relevant causal relations). And the RT effectively clarifies what it is that makes it the case that the specified pattern of causal relations, rather than some other, constitutes M. Then, for a given instantiation of the mental state in question, the physicalist vertical explanation would consist in citing a set of instantiated physically-based causal relations and showing, by appeal to the functionalist RT, why those relations suffice for the realization of that mental state.26

It is important to understand that an RT for an attribute N is, in most cases, an abstract theory that, while specifying what it takes to instantiate N, does not normally make any reference to physical attributes. Rather, an RT specifies something like the essence of the attribute, although I don't wish to hang too much on the use of that term.27 It delineates a configuration of attributes that suffices to instantiate N in any possible world, not just physically possible worlds. Thus, pursuing the example further, functionalism is compatible with dualism, although it is also clear why physicalists can appropriate a functionalist view of the mind to provide accounts of how mental attributes are realized by physical systems. According to the functionalist, the essence of a mental state, M, is a certain pattern of causal relations. Such patterns can be exhibited in a physically-based world of the sort envisioned (p.212) by the physicalist, or such patterns can be exhibited in non-physically-based worlds (for example, spirit worlds).

Further, and to switch examples, transparency can be instantiated in worlds without physical microstructures like those that underlie transparency in our world. Just so long as something is capable of being seen through, then it is transparent. Our physics provides an account of the physically-based structures that allow for the instantiation of this attribute in our world. But we should not confuse a parochial, physical account of transparency with its essence, just as we should not confuse a parochial, physically-based account of mentality (for example, in terms of neural structure and processing) with its essence. Such attributes as transparency and mentality can be instantiated in worlds that, although possibly similar to our own at macroscopic and phenomenological levels, are very different at microscopic levels. It is also evident that not every attribute is, or even can be, instantiated in every world. Thus the value of RTs lies both in revealing why it is that an attribute is or can be instantiated in a given world and in providing a way of understanding why it is that an attribute is not, and perhaps cannot be, instantiated in a given world (i.e. none of the possible configurations of attributes in that world suffice for realizing the conditions specified by the RT for that attribute).

Another feature of RTs is that there is not just one sort of RT appropriate for all attributes. Given the many different types of attribute that there are (for instance, intrinsic, extrinsic, causal, structural), there will inevitably be variation in both the kinds of attributes cited by RTs as well as the ways in which attributes combine to form an instantiation of the target attribute. In particular, since some attributes are intrinsic while others are extrinsic, it should be underscored that the realization base for an attribute need not be localized in the same region as the attribute that is realized. Whether or not a given attribute is instantiated may well depend upon what is going on elsewhere in the world. And whereas higher-level causal powers are, perhaps, constituted by the combining of lower-level causal powers, various more abstract features of things (for instance, semantic properties) will depend upon non-causal properties and relations (for instance, as in Post's Ammon example discussed earlier).

RTs provide conceptual structures for understanding how non-physical phenomena can be realized in a world. For the physicalist, (p.213) they abstractly characterize the kinds of physically-based structures that suffice for realizing non-physical attributes and that provide physically-based models for higher-level theories of this world. As a consequence of this abstractness, they allow us to see how it is possible for there not to be any lower-level systematicity in the class of realizations for an attribute: what the members of such a class share is abstract relative to lower-level theory. On the other hand, such RTs help to account for significant relations of covariation between lower- and higher-level domains (for example, relations of determination or nomological correlation). They flesh out the fabric of the connections between domains, and in so doing they allow us to distinguish between those connections that are instances of realization relations and those that are only accidental or only nomologically correlated with relations of deep dependence.

As a consequence, RTs reveal how the physical bases ground all that takes place in our world, thereby helping to provide a framework for explanatory unity, integration, and connection between everything that can or does occur. This includes, remarkably, the reining in to the physicalist fold of even private or otherwise inaccessible phenomena. To account for how such phenomena can be realized by physically-based phenomena is not to undermine their privacy or inaccessibility. Knowing the character of the physical basis of, say, private mental states possessing qualitative attributes is not to know everything that is true of such states. But this does not warrant the conclusion that physicalism is thereby refuted. To explain why a certain attribute is instantiated is not to possess the attribute, and thus it is not to come to know those things that only a subject who possesses the attribute could know. That some (for example, Jackson 1986; Nagel 1974) nave tried to parley such obvious truths into an objection is a mistake based on a faulty understanding of physicalism, in my opinion (see P. M. Churchland 1985 and Van Gulick 1985 for discussion).

It should be clear from the above discussion that RTs are not explicit, physical definitions of non-physical attributes; nor are they identity statements, statements regarding physical/non-physical determination relations, bridge laws, or connective principles with some weaker modal force. In fact, they need not explicitly concern vertical, physical/non-physical relations of any sort. Rather, RTs are, in general, more abstract than any of these other sorts of (p.214) principle. It is only when they are applied, as the physicalist requires, that they begin to clarify relations between physical and non-physical domains.28

It is especially important to see that RTs are not just explicit definitions in physical terms, as earlier versions of the programme had it. Even if physical definitions were forthcoming, there would still be a need for RTs if the physicalist goals regarding explanation are to be served. That is, there would still be pressure to understand why the instantiation of the members of a given set of physical attributes realizes a given non-physical attribute, even if the set of physical attributes can be converted into a definition of the non-physical attribute.29 It is, in my opinion, a major shortcoming of both classical and recent versions of physicalism that the role of explicit definitions has not been very well understood in terms of the goals of the programme.

Causey (1977), for example, is at great pains to distinguish those nomological correlations which express attribute identities from those which express causal relations. His criterion for making this distinction is the existence, or lack thereof, of an explanatory account of a causal relation between the attributes in question. On my view, Causey's discussion is faulty because his dichotomy is not exhaustive. He leaves out the possibility of the realization of one attribute by another and he takes much too seriously the idea of attribute identity across domains (it is a rarity at best).30 He has the right idea in seeking explanatory accounts, but he encourages a false picture of the inter-domain problem of how attributes in one domain relate to attributes in others. The cost is a failure to appreciate the need for inter-domain explanations in general. As a consequence, the goals of the physicalist unification programme are sacrificed by an inappropriate emphasis on identity. (p.215)

Further, as pointed out in Chapter 2, the view in Friedman (1975) that each non-physical predicate is associated with a (possibly infinite) class of physical predicates, although sufficient for his purpose of characterizing ‘weak reduction’, raises, but does not indicate how to answer, the question, ‘Given a non-physical predicate, in virtue of what is each member of the associated class of physical predicates a member of that class?’. Simply to say that each expresses an attribute that is a realization of the attribute expressed by the given non-physical predicate begs the explanatory question. At the heart of that question is the quest for an understanding of why certain physical attributes realize a given non-physical attribute while other physical attributes do not. To respond only by asserting that the former attributes do while the latter do not begs the question and leaves unaddressed a large host of mysteries concerning realization. The point of T4 is to rid a physicalist system of such mysteries.

The general idea is that it does not matter for the achievement of the explanatory goals of the programme whether each non-physical attribute is associated with a single coextensive physical (or physically-based) attribute or with a class of such attributes each member of which is sufficient for the realization of the non-physical attribute or with a class of classes of such attributes where the members of each class jointly provide a realization of the non-physical attribute. Either way, the explanation of the realization of non-physical attributes by physical attributes is required if a large class of vertical mysteries are to be eliminated from a physicalist representation of nature. The role of realization theories is to provide a basis for the elimination of those mysteries. To satisfy the explanatory motivations underlying the physicalist programme, T4 requires explanations, via RTs, of the realization of all non-physical attributes by associated physical (or physically-based) attributes.

RTs are also not the same as what Post refers to as ‘connective theories’, which are largely empirical theories of the specific connective relations that hold between physical and non-physical attributes in this world.31 Both sorts of theory are, in my opinion, essential to the physicalist system and they should be viewed as complementing each other. Whereas RTs provide the more abstract (p.216) account of how attributes can be realized, an account that provides guidelines for the study of specific realizations of attributes, connective theories provide the specific answers to questions concerning how non-physical attributes are in fact realized in this world. In terms of the quest for vertical explanations of the instantiations of non-physical phenomena, RTs provide the general guidelines for how to isolate and combine relevant attributes bearing upon some specific phenomenon under investigation, whereas connective theories make the specific identifications of physically-based attributes regarding a particular instantiation, or class of instantiations, of that phenomenon.

To summarize: an RT provides an abstract account of what it is in virtue of which a non-physical attribute can be instantiated (i.e. an abstract characterization of the types of structures and processes that suffice for constitution of the attribute). The provision of RTs and the provision of vertical explanations of specific realizations of attributes drawing upon such RTs is one of the goals of research programmes concerned with understanding why higher-level attributes are realized by lower-level attributes. And the attainment of such goals relative to physically based attributes furthers the more general goals of the physicalist programme outlined in Chapter 1 by promoting the unification of knowledge via vertical explanatory connections, and thus by increasing understanding.

Comparable points can be made regarding the realization of objects in nature, T4 calls for the vertical explanation, in terms of RTs, of all particular instantiations of the various non-physical kinds of individual object. The RTs provide the general account of what it takes to realize a given kind of object, and particular vertical explanations employ the RTs to identify the relevant, physically-based objects and attributes that combine on a given occasion to constitute, and hence realize, a particular individual of the given kind. The result is an understanding of how individual objects are embedded in and realized by the physically-based fabric of nature. I shall bypass further discussion of this aspect of T4 for reasons of space.

Turning now to a discussion of the thesis concerning physicalist explanations of regularities in nature, early versions of the physicalist programme were concerned with what was called ‘the unity of laws’ in science (Carnap 1969). The idea was, simply, that all (p.217) laws of nature are derivable from the laws of physics plus suitable bridge principles. Such derivations were thought to be sufficient for the explanation of those laws and of the regularities they expressed. Although I fully support the motivations behind this version of the programme, I am sceptical about the specific thesis proposed to serve them. That is, I fully endorse the idea that the physicalist programme is motivated by a quest for increasing the explanatory power of a system of knowledge by, inter alia, vertical explanation of regularities, but I reject the classical thesis that expressed this motive. The task now is to formulate an alternative.

The physicalist's conviction is that all regularities, their instances, and their exceptions are determined by underlying physical phenomena. This conviction was captured earlier, from an ontological point of view, by the thesis, T1, concerning the realization of objects and attributes, and the thesis, T2, concerning the physical determination of fact. The question now is: ‘How is this conviction to be expressed from the point of view of a system of knowledge concerned with those facts?’ To this end, I shall focus attention on the following question: ‘Given a set of non-physical regularities characteristic of some domain, what are the underlying physically-based structures and regularities that realize them?’ It is this and closely related questions that guide inquiry within inter-field disciplines (for example, physical chemistry, microbiology, the neurosciences). In short, vertical inquiry is concerned with the study of lower-level ‘mechanisms’ that underlie higher-level regularities.32

Although inter-field inquiry (Darden and Maull 1977) aims at the identification of mechanisms for regularities, the questions addressed by inter-field research programmes make no presumption that higher-level phenomena are uniquely realized by lower-level phenomena. Thus no presumption is made that exactly one lower-level ‘mechanism’ underlies a given higher-level regularity; rather, there may be different mechanisms that realize its various instances. As a consequence, the explanation of a regularity should be understood as the provision of members of a class of mechanisms (p.218) that jointly suffice to account for all the possible instances of that regularity. Given the potential openendedness of such a class, there is no assurance that all the members will, or even can, be identified as a consequence of human research efforts.

It should be further kept in mind that, although the physicalist is ultimately concerned that there exist fundamental physical structures and processes that underlie regularities, it is allowed that there may well be many intermediate levels of organization, complexity, and abstraction.33 For the purpose of explaining a given non-physical regularity exhibited by a given sort of system, the underlying mechanisms may themselves be non-physical in character and they may be constituted by objects and attributes drawn from several different levels of organization of the system. As emphasized in Chapter 1, the physicalist's only requirement is that any sequence of levels of realization of an instance of a regularity is ultimately grounded in the physical basis.

Alternatively put, one aim of inter-field research is to identify physically-based ‘models’ within which to embed higher-level phenomena. But simple embedding is not enough; an account of how lower-level, physically-based structures and processes suffice to realize instances of the various regularities exhibited at higher levels is required as well. And, as pointed out earlier in connection with T4, such explanation is not explanation in physics or any other lower-level discipline per se. It is a sui generis form of explanation that accounts for the vertical links between lower-level and higher-level domains. Inter-field inquiry is distinct from, although highly pertinent to, the fields that it aims to connect.

Given these preliminaries, the physicalist thesis concerning the explanation of regularities is as follows:

  1. (T5) All instances of, and all exceptions to, natural regularities are explainable in terms of physically-based phenomena.

The idea is that the physicalist requires no more than that every possible instance and every possible exception to the regularities holding at higher levels are explainable in terms of lower-level, physical or physically-based phenomena. Such explanations consist (p.219) in identifying the non-physical attributes involved in the regularity to be explained, identifying the relevant physically-based attributes involved in the specific instance that is to be explained, and providing an account of the relations between the physically-based attributes in question which constitute the causal or non-causal relation between the non-physical attributes.

For instances of some non-physical regularity, if an RT for each attribute involved is known, and if, as a result, the physically-based attributes that realize the non-physical attributes can be identified, and if relevant physically-based relations between those attributes are known, then an instance of a regularity is explained if the relation between the non-physical attributes is explained in terms of the relations between the physically-based attributes that realize the non-physical attributes. That is, if it is a high-level regularity that non-physical attribute A is a cause of non-physical attribute B, then an instance of this regularity is explained by citing the physical attributes C and D, which realize A and B respectively on that occasion,34 and by explaining that C causes D. The physicalist thesis is that every possible instance of a non-physical regularity can be explained in roughly this way, although it is not assumed that all relations of interest are causal.

The idea that regularities studied in higher-level disciplines typically have exceptions is an important one whose significance for the concept of a law of nature has not been adequately studied. That all regularities in nature are exceptionless is untenable because, as one ascends a hierarchy of phenomena ordered in terms of complexity and the relation of realization, regularities at higher levels will typically have exceptions due to interferences from outside of a system, ‘intrusions from beneath’ (for example, defects in a mechanism), or failure of satisfaction of certain other ceteris paribus provisos (i.e. conditions required for the regularity to hold).35 In my development of the physicalist doctrine, I assume that there can be such exceptions and that they must properly fit into a physicalist framework.36 (p.220)

Specifically, the physicalist requires only that such exceptions be explainable along the same lines as the instances. For example, if an exception to a causal regularity is a case in which the antecedent, but not the consequent, attribute is instantiated, some other attribute being instantiated instead, then such an exception is explained via an account which clarifies the relations between the attribute(s) realizing the antecedent non-physical attribute and the attribute(s) realizing the non-physical attribute that replaced the expected consequent non-physical attribute and which explains how those relations between the physically-based attributes realize the relation between the two actually occurring non-physical attributes. In actual contexts of explanation, such vertical explanations will be embedded within a larger framework that explicitly reveals what particular intrusion, breakdown, or violation of ceteris paribus conditions led to the exception. Thus the explanation shows why the antecedent attribute was not accompanied by the attribute specified by the regularity, but was accompanied by the unexpected attribute instead. The physicalist view is that every possible exception to a non-physical regularity has such an explanation.

This aspect of the physicalist programme is, in my view, especially important. To acknowledge and take into account the existence of exceptions to high-level regularities, and hence the degrees of lawlikeness that high-level generalizations can possess, is to recognize how contingent, high-level structures can exhibit regularities that are both counter-factual supporting and yet not strict. It is a virtue of physicalism that it aims to account for these features of high-level systems. The counter-factual force of such regularities is based upon and realized by lower-level, relatively more exception-free physically-based regularities. The exceptions are explained in terms of intrusions, breakdowns, and inappropriate background conditions which are also grounded in the physical basis.37 The embedding of a system in a larger physically-based (p.221) structure and the vertical explanation of the attributes involved provides an appropriate framework for such explanations.

Such physically-based embedding has an additional virtue. With the study of breakdowns in the microstructure of a system and with the study of the systematic and non-systematic reactions of such microstructure to external influences, a deepened understanding of how a macro-level system works is gained. In particular, macro-level breakdown patterns can be better understood and dealt with. It is not just that the physicalist wants exceptions to be explained. The objective here is, again, increased understanding of how the system works. To ignore the non-exceptionless character of high-level regularities is to undermine such increases in understanding.

In light of this, it is easy to see that the emphasis some philosophers (for instance, P. M. Churchland 1984; P. S. Churchland 1986) place upon the methodological importance of ‘smooth reductions’ of one theory to another is misleading. The idea that if a theory is not smoothly reducible to lower-level, physically-based theories then it ought to be eliminated from science is flawed inasmuch as it does not take into account how exceptions to the higher-level theories arise. In my opinion, many of the arguments offered by the Churchlands, for example, in support of so-called ‘eliminative materialism’ are facile in this regard. Put another way, too aggressive an emphasis upon growth reductions, where the aim is to weed out the errors embodied in older theories by developing more powerful and more accurate newer theories to subsume and replace them, can lead one to mistake an important exception to a generalization for an error in theory. Structural reductions, in which the aim is to embed higher-level systems and theories within lower-level systems and theories, are geared specifically to accommodate exceptions which reveal the deeper character of what is going on. An exception to a law is not necessarily indicative of a fault. As a result, to aim singlemindedly for a system of exceptionless generalizations is very likely going to lead to missing what is going on in this regard. Although progress in science requires the elimination of error, this should not be equated with the elimination of non-exceptionless generalizations. Rather, (p.222) the possibility of systematic or non-systematic exceptions must be accommodated and accounted for if high-level systems are to be adequately understood.38

As a consequence of these points about the sources of exceptions as well as points regarding the multiple realizability of high-level phenomena, T5 does not imply the derivability of all non-physical laws from physical laws. Hence it does not appear to capture the physicalist idea that the non-physical laws are explainable in terms of the physical laws in the way that the classical unity of laws thesis supposed. However, the derivability of laws is not required to capture the physicalist idea that everything that does or could happen in nature is explainable in physical terms. T4 and T5 do not leave room for something to take place without there being a physically-based explanation for it. What may go unexplained are the regularities themselves, if there is something more to explaining a regularity than explaining all of its instances. This is not a major loss for the physicalist as long as everything that either can or does take place in nature admits of a physically-based explanation. Perhaps it should have been expected that if not all attributes are physical attributes and if non-physical attributes can be multiply realized by physical attributes, then the regularities involving non-physical attributes would not in general be derivable from physical regularities.39 And, as has recently been pointed out, the existence of genuine exceptions to higher-level regularities makes it undesirable to have derivations of such ‘laws’ from true, lower-level generalizations that are free of exceptions. Physicalists should therefore reject the classical derivational thesis as well as the thesis that all non-physical regularities are explainable in terms of physical laws, if this latter thesis requires such derivations. In their stead, T5 offers as much as the physicalist needs in order to achieve the explanatory goals of the programme. And it does so in a way that handles the problem that exceptions pose to the classical unity of laws.40 (p.223)

It should be pointed out that T5 does not preclude the possibility of finding unique underlying physical mechanisms for some non-physical regularities or of discovering exceptionless non-physical regularities. Nor does it preclude finding derivations of non-physical laws from physical laws. The point is, rather, that these are not requirements of the programme, since the explanatory goals of physicalism are served fully by T4 and T5. The aims of vertical explanation are to identify mechanisms that can individually play a role in accounting for some (or all) of the instances of a given regularity and that can jointly account for them all, and to identify the sources of whatever exceptions the regularity might exhibit. These aims are served whether there is one or many such mechanisms.

A further virtue of T5 is that it allows for the existence of ‘emergent’ laws in at least two senses. First, there may be laws that are not derivable, hence not predictable, on the basis of knowledge of lower-level laws. Second, there may be laws that ‘kick in’ at certain stages of development of the universe. RTs and vertical explanations based upon them will allow us to understand why certain phenomena occur and certain regularities hold only when certain kinds of structures have evolved. Neither of these sorts of emergent phenomena falsify physicalist theses, of course.

With respect to the integration of theory, T4 and T5 serve physicalist purposes in a number of ways. The existence of a system of explanations that is physically-based means that all explanations of phenomena can be linked, if necessary, by digging as deeply and as widely as is required to discover a physically-based ground for such linkages. Such a ground allows us to relate different ways of understanding how a system works (for example, as in cognitive science where psychology, linguistics, artificial intelligence research, anthropology, etc. are integrated in the study of cognition). And such a ground provides a framework for relating various theories in different domains (as when our understanding of economic factors is related to our understanding of social forces and events). Links between domains, especially links of a causal nature, are underwritten by physically-based relations and explanations. As I emphasized in Chapter 1, the physicalist maintains that all paths of interconnection and influence must be explainable in terms of physically-based phenomena.

It follows that the issue of the autonomy of domains of knowledge (p.224) is addressed within a physicalist framework by denying that there are any distinct domains that are absolutely autonomous. Every domain is related to physics in ways prescribed by the theses (i.e. they are related in terms of relations of determination, physically-based realization, and physically-based vertical explanation). And every domain is related to every other domain via the common physical basis. The details of the relations between domains will vary from case to case, there being stronger constraint relations in some cases than in others. But since the collections of physically-based models and associated explanations that exist for various domains of inquiry are drawn from a unified, common, physically-based pool, it is not possible for any domain to be entirely independent of the others. In this regard, it is especially important to note that physics is not an autonomous domain in so far as physicalism requires that it should bear certain relations to all other domains.

Finally, to return to the issue of how distributions of objects and attributes in a higher-level, non-physical ontological domain is constrained by the ‘laws’ characteristic of that domain, we observed earlier that, since such laws are not exceptionless, the constraints are more complicated than in the case of the physical domain. Given our discussion of such exceptions and how they are to be accommodated within a physicalist system, the main constraint on such distributions is that either they are in conformity with the counter-factual supporting generalizations of the domain or they are in conformity with deeper and more comprehensive generalizations of greater nomic force. That is, any exception to a generalization must be subsumed under some more exception-free generalizations that rein in the sources of the exception.

This completes my discussion of the physicalist theses. In the next two chapters I shall present and discuss a set of metatheses concerning them and I shall frame and evaluate a number of outstanding objections against them to which physicalists ought to provide satisfactory responses.


(1) As an expository convenience, I shall speak of the ‘instantiation’ of objects in regions of space-time, although this usage is non-standard.

(2) Although I deny the above-mentioned identity theses for attributes, I allow that type or token identity theses for other ontological categories may be plausible.

(3) A region R is identical to a region S just in case R and S contain the same space-time points. Since such regions need not be connected, but can, instead, be scattered over space-time, objects too can be scattered over space-time.

(4) Identification of the objects, attributes, and operations is a matter for research within each domain.

(5) Post has suggested (in correspondence) that there are apparent exceptions to this thesis that arise in the context of fundamental physical theory: e.g. (1) singularities that are not in space-time but which are boundaries of space-time (‘real enough but of less than four dimensions’), and (2) more fundamental (i.e. less structured) spaces out of which space-time structures originate—such spaces are neither space-times nor in space-time. If either of (1) or (2) is real, then minor adjustments in To are required. However, neither seriously compromises the spirit of the thesis, in my opinion.

(6) In the case of causal properties, such a restriction may be required. But the physicalist need not be bound to this demand with respect to all attributes.

(7) Note that, consistent with the discussion in Ch. 2, I make no requirement either that the attributes in P involve at least some causal relations or that they are instantiated in the same region as the attribute N. Attributes that bear upon the realization of N can be instantiated in regions distant from the region in which N is instantiated. For example, the attribute of being the tallest mountain is realized by relation to other mountains which are both at a distance and not causally related to the tallest mountain in any significant way.

(8) A reductive thesis, as I understand the concept, is one which relates the members of one domain to the members of another, subject to certain constraints. Thus there can be many different sorts of reductive claims depending upon the domains involved and the kinds of constraints required. A non-reductive thesis, on the other hand, requires only that two domains be related in some global, systematic way without stipulating relations between the individual members of the domains.

(9) It is the job of what I shall call ‘realization theories’ to clarify, for each kind of object, how these uncertainties are to be managed.

(10) Although I doubt whether either of these is a plausible example of identity.

(11) Strictly speaking, Quine holds that there are behavioural facts of the matter, but these behavioural facts do not suffice for fixing a uniquely correct translation. It is the residual slack that makes room for subjective variation.

(12) A complete distribution is one that is based upon an evaluation of every physical attribute and every physical kind of object at every region of space-time. I take this to be compatible with the idea that some assignments cannot be specifically made on such an evaluation (e.g. some magnitudes have indefinite values at certain regions). Instead, the evaluation may assign a probability distribution for a range of values of that magnitude.

(13) The idea behind this strategy is due to Hellman and Thompson (1975), although I depart from their approach.

(14) This does not assume that all laws of nature are exceptionless. This class will probably include states of nature that are in violation of certain ‘laws’, although compatible with other, more fundamental, laws. See below, in my discussion of T5, for further elaboration of this point.

(15) T2 is pretty nearly equivalent to the explications of truth determination found in Hellman and Thompson (1975: 558, #4), and Post (1987: 185, TT*; 1991: 118, D).

(16) This, of course, is exactly the Quinean position regarding the attributes studied in semantics, mentalistic psychology, and linguistics.

(17) Henceforth, I shall refer to the conjunction of the theses T1a and T1b simply as ‘T1’.

(18) See Post (1987: ch. 2.) for discussion of ‘ultimate explanation’.

(19) E.g. whereas their formulation relies upon the model-theoretic notion of elementary equivalence, T3 does not. Since elementarily equivalent models need not agree on higher-order truths, it is possible that their formulation does not capture the full thrust of physicalist truth determination.

(20) To say that all instantiations of non-physical objects and attributes are explainable does not imply that such explanations will inevitably be developed or even that such explanations are necessarily within our cognitive reach.

(21) For ease of discussion, I shall focus on attributes, although the points I shall make apply to objects as well.

(22) Although I am emphasizing the idea that attributes in the explanans of a vertical explanation are drawn from lower levels, I do not mean to rule out the possibility that a vertical explanation might appeal to a physically-based attribute at the same or higher level as the attribute whose realization is being explained. The central requirements are that all attributes in the explanans are physically based, thus ensuring that the target attribute is ultimately related to the physical domain, and that an explanation is provided of how that target attribute is realized on the given occasion.

(23) It is a red-herring issue, sometimes raised against any form of physicalism which introduces connective principles that are not part of physics, that somehow the priority and privilege of physics are compromised by appeal to such principles. Such an objection betrays a misunderstanding of what physicalism involves and of how vertical explanation works.

(24) Nor do I make any assumption that there is transitivity of vertical explanation in the sense that if there is a vertical explanation of some attribute, N, in terms of physically-based attributes, then there must also be a vertical explanation of N in terms of exclusively physical attributes. See Post (unpublished) for discussion of this issue.

(25) Similarly, a realization theory for a kind of object, O, reveals what sorts of configurations of physically-based objects and attributes constitute instances of O.

(26) No endorsement of functionalism in the philosophy of mind is intended here.

(27) I want to leave open the question of whether there could be more than one RT associated with a given attribute. Although it is clear that, relative to a single RT, multiple realization is possible (e.g. as in the case of functionalism), it is much less clear whether a given attribute can have more than one RT. See Cummins (1989) for suggestive discussion of this issue with respect to mental representation.

(28) Although I think Putnam (1979: 609) is right in insisting that any form of physicalism must involve ‘type-type’ relations of some sort, he does not get the point quite right since he seems to have in mind physical/non-physical correlations.

(29) Note that this assumes that attributes are not individuated by the criterion of nomological coextensiveness which is the standard constraint upon physical definitions. If this were the individuation criterion, then definability would lead directly to identification of the attributes expressed by the defined and defining terms.

(30) The standard examples of theoretical identities, such as those involving temperature or pressure and statistical mechanical properties of an aggregate of molecules, are quite flawed, as Earman suggests in the quotation at the beginning of this chapter.

(31) See Post (1987: ch. 5) for valuable discussion of connective theories and the roles they play in a full development of a physicalist system.

(32) The use of the term ‘mechanism’ here should not mislead one to think that the brand of physicalism I am discussing is a species of 19th-century ‘mechanism’. The term is simply a stand-in for whatever physically-based phenomena constitute the realization of a regularity in nature. Thus by ‘a mechanism for a regularity’ I mean a system of physically-based objects and attributes governed by a class of physically-based regularities.

(33) The neurosciences provide an excellent illustration of how many different levels of organization and function are concurrently studied for the purpose of identifying mechanisms for the various regularities that are exhibited by the human nervous system.

(34) Among other things, C and D constitute the causal powers of A and B on this occasion.

(35) See Fodor (1975) for preliminary discussion of this neglected topic. And see Schiffer (1991) and Fodor (1991) for recent discussion of some of the issues as they bear upon psychological ‘laws’.

(36) Field (unpublished) has also recently emphasized the importance of this demand on physicalism.

(37) A generalization will have diminished lawlikeness to the extent that it fails to take into account all the factors that could influence the phenomena it concerns. Such a failure is not necessarily a shortcoming, however. To the extent that there are unsystematic causes of exceptions, it would often be impractical and futile to try to incorporate all possible unsystematic influences in the statement of a law. On the other hand, idealizations away from systematic causes of exceptions can be of considerable value in gaining understanding of certain aspects of how a system works. Since the attainment of exceptionless generalizations at high levels would require the incorporation of all systematic and unsystematic causes of breakdown, such generalizations are often either unattainable or undesirable. Thus, for both practical and theoretical reasons, the existence of exceptions need not undermine the importance of generalization.

(38) I do not wish to be seen as opposing the value of growth reductions (in an appropriate sense of ‘reduction’). What is required is recognition of both sorts of reduction, and attainment of a proper balance of the two in methodological arguments.

(39) See Fodor (1975) for a version of this sort of argument.

(40) See Field (unpublished) for a different approach to the explanation of high-level laws.