• No results found

The non-mereological candidate relations

In document Making sense of response dependence (Page 66-71)

1. What is the relation between levels?

1.1 The non-mereological candidate relations

In the following, I shall risk a brief account of the non-mereological relations and the differences between them, though it must necessarily be superficial, and may be controversial. The task is complicated by the fact that terms like realisation, constitution, and possibly even supervenience, do not refer to well understood phenomena, but tend to function as place-holders for relations that are ubiquitous yet not well understood – identity-like relations that are different from identity.

Supervenience, realisation, and constitution have one thing in common that sets them apart from the part/whole-relation. In order to determine what things stand in the part/whole relation to each other, we need only look at the actual facts in a certain spatial region. By contrast, in order to determine whether two phenomena supervene on, realise, or constitute each other, we must take into account how things are in other possible worlds and/or at different times and places in the actual world; variation across worlds and situations is a tell-tale sign that the relevant relation is not identity, but one of the ‘identity-like’ relations of realisation, constitution or supervenience.

Supervenience, realisation and constitution differ from each other in that constitution is a relation betweenobjects, while realisation is normally thought of as

a relation between properties, and supervenience relations can hold between

properties or states of affairs.

Constitution, as traditionally understood, denotes a relation between objects that occupy the same spatio-temporal region, but differ in their modal or past or future properties. Classical examples are a statue and the clay it is made of, or a cat

and its body, or a ship and the sum of materials it is built of. For practical purposes, the relata of the constitution relation count as the same object, but due to their difference in modal properties, they cannot be identical (at least not in the ordinary, strict sense, though some authors allow for a less strict notion of contingent identity5).

The notion of realisation is part of a functionalist two-level story involving role properties and realiser properties. A role property is normally understood to be the (second order) property of having a (first order) property or properties playing a certain, often highly complex, causal role. Realisers are the properties that occupy the roles and do the causal work in question.

Thus, the core claim of functionalism involves two ‘levels’, or ‘orders’, of very different properties.

The picture might perhaps be extended to include further levels – either by regarding some roles as realised by properties that are themselves realised by further properties, or by having several levels of role properties that are all realised by a basic level of realisers (though this version would require a story about why it makes sense to think of the role properties as arranged in a hierarchy of levels).6

Presumably the functionalist apparatus can be extended to cover roles that are not defined in purely causal terms, though this would change the spirit of the proposal as described above, and would require an account of how role and realiser properties are related if not by realisers filling causal role-descriptions. Also, it might be extended to include non-physical realiser properties (raising the same problems).7

5

See e.g. Yablo (1987).

6

However, the extension to further levels sits uneasily with core thoughts behind functionalism. The traditional functionalist story is an attempt to tie higher level properties to more physicalistically kosher properties, and properties the appeal to which can save intuitions about causal efficacy in the face of causal exclusion arguments. On this approach, higher and lower level properties are very different kinds of properties. Hence, to reiterate the story for more levels than two would presumably not preserve the original motivations (especially not if the assumption that there is a basic level must be relinquished; see section 4).

7

Supervenience is probably the relation between levels that is most central to this work. In the following, whenever I speak of levels and it is not clear from the context what sense I am talking about, it is safe to assume that it is the supervenience sense. One reason that supervenience is so well suited to capture the relation between levels in the sense most interesting for present purposes is that it comes in a wide variety of versions, sufficiently similar to say that we’re talking about levels in a roughly unified sense, and yet sufficiently different to allow for the diversity of the issues under consideration. The core idea is that a group of (high level) properties, let’s call them the B-properties, supervene on another group of (low level) properties, the A-properties, iff no two possible situations are identical with respect to their A- properties while differing in their B-properties.8 Supervenience claims come in local and global forms, depending on how broadly the ‘situations’ in question are understood. If the claim is that the A-properties ofindividuals(in a sufficiently broad sense of the term that includes objects, people, societies etc.) determine the B- properties of those individuals, we get local supervenience. If the supervenience claim concerns the distribution of A- and B-properties inentire worlds, we getglobal

supervenience. Local supervenience entails global supervenience, but not the other way around (intuitively, if the B-properties of each individual could not be different without a difference in the A-properties of that individual, then the B-properties of all the individuals taken together cannot be different without a difference in the A- properties of all the individuals). On another axis of variation, the ‘possible’ in the supervenience schema can be filled out in several ways, depending on what worlds are considered relevant. It can be logical possibility, in which case we get logical supervenience. Or it can mean nomic or natural possibility, yielding nomic or natural supervenience – supervenience in all worlds where the laws of nature are the same as in the actual world. Thirdly, it can mean metaphysical possibility, yielding

metaphysical supervenience.9 The distinction between logical, natural and

metaphysical supervenience is independent of the one between local and global

8This formulation is based on Chalmers (1996) p. 33; there are other ways to do it, but for

simplicity we shall stick with this version.

9

Some, e.g. Chalmers (1996), doubt that there is such a third category, while others take it for granted.

supervenience; the versions can be combined as you wish. Besides these most central varieties, there are others. One is ‘weak supervenience’ which says merely that in the actual world, a difference in B-properties always comes with a difference in A- properties (i.e. a local supervenience claim restricted to the actual world; the global version would make little sense). Another is Blackburn’s ‘ban on mixed possible worlds’, which claims local supervenience of B-facts on A-facts within each possible world, but not across worlds.10 It is possible to design supervenience claims to suit a wide variety of purposes.

Though it has received surprisingly little attention in the literature, there is also scope for variation regarding the source of the supervenience – e.g. whether it is underwritten by requirements on how the concepts function (as in ethics), or by relations that obtain between thereferentsof the concepts.

Which versions of supervenience are relevant to which domains? Are e.g. biological or economical facts logically supervenient on physical ones, or only naturally or metaphysically supervenient? In which sense, if any, do mental facts supervene on physical ones? These are hard questions and provoke a lot of disagreement. Hence, there is lots of work to be done in deciding which supervenience relations (if any) are appropriate with which level-differences. Of course, one is free to combine different supervenience claims for different domains. For example, one can hold (like Chalmers 1996) that biological properties supervene logically on physical properties, whereas phenomenal properties supervene only naturally. Which version of supervenience is relevant is thus not a matter to be decided once and for all.11

A supervenience claim is not the final word on the relation between two kinds of properties or states of affairs. It says something about correlation patterns between

10Blackburn (1984), Ch. 6.

11 Something similar is true when supervenience is combined with response-dependence;

different supervenience relations may go with (and possibly even distinguish or define) different versions of response-dependence, depending of the choice of supervenience bases and the class of worlds quantified over. Unfortunately, the relationship between response-dependence and supervenience will not be investigated (much) further in this work. To carry out such an investigation properly would require detailed investigation into particular domains, and this falls outside the scope

them, but it says nothing about why, or in virtue of what, the correlation pattern is as it is. In that sense, supervenience claims state problems rather than providing solutions. This, of course, is a valuable contribution in itself. But it means that, like ‘realisation’ or ‘constitution’, ‘supervenience’ is a sort of place-holder for the explanation we seek, but is not the explanation itself.

Finally, identity – the main rival of the idea of ontological level differences, and the relation that will be relevant if level differences are merely differences in mode of presentation.

In David Lewis’s terms, the identity relation is easy to explain: everything is identical with itself and not with anything else. The relation is more fully characterised by Leibniz’s law: If A is identical with B, A and B share all their properties. Some also endorse the principle of the identity of indiscernibles: If A and B share all their properties, then A is identical with B.

Some authors hold that identity can be relativised to times, or to concepts /sortals, or to worlds. This means restricting the range of properties that A and B must share if they are to count as identical. For tensed identity, A and B may be identical at t even if they differ in their properties at other times. For contingent

identity, A and B may be identical even if they differ in other possible worlds. For

sortal relativity, the idea is that A, say a statue, and B, say the lump of material it is made of, can be the same material object but not the same work of art (i.e. share all properties relevant to the first sortal but not all those relevant to the second).

With respect to the levels framework, the distinction between contingent identity and strict identity is especially relevant. The traditional view is that if A and B are identical, they are so necessarily. The reasoning is as follows: Anything is necessarily identical with itself. So, in particular, necessarily (A=A). If A is identical to B, by Leibniz’ law, they share all their properties. One of A’s properties is this: necessarily ( _ = A). So B, too, must have this property, and hence be necessarily identical to A. But if this is the case, then e.g. statues and the lumps of material they are made of can’t be identical, since they can come apart in other possible worlds. They might, instead be viewed as objects at different levels, related by the constitution relation. Or one could argue that contingent identity is possible, and relevant to such cases. Contingent identity might be compatible with level

differences in a certain, fairly weak sense. But identity in the strict sense is not a relation between different levels; it is the negation of the claim that there are level differences.

In document Making sense of response dependence (Page 66-71)