1. There are at least three important questions to be dealt with in this essay. They are:

  1. What are the criteria of existence?
  2. What exists?
  3. What are the criteria of ontological commitment?

2. To ask "What are the criteria of existence?" is to ask how it is that we should go about finding the answer to existence questions such as "Is there a swimming pool at Keele?", "Is there a prime number between 4799 and 48037?", "Does Uranus exist?", "Do numbers exist?" and many others. The criteria of existence which we normally use vary widely depending on the sort of thing whose existence is in dispute. Some of the existence questions concern the existence of individuals, and some concern the existence of classes of individuals. For this reason existence questions form a hierarchy. Every existence question is logically related to a whole series of other existence questions. The existence of 'material objects' for example, is a necessary condition of the existence of protons. More generally, the existence of any individual or class of individuals is logically dependent upon the existence of any class which contains that individual or class of individuals.

3. It may be objected that the logical dependency, if there is one, must be the other way round. The existence of an individual being a necessary condition for the existence of any class of which it is a member. In deference to this objection I shall try to express myself more precisely. We are concerned with hypothetical entities, and with the ways in which we establish which of these hypothetical entities actually exist. The classes which concern us are classes of these hypothetical entities, and I shall presume that the existence of these classes is independent of the existence or otherwise of the hypothetical entities which are its members. What concerns us about the classes is whether their membership is reasonably well defined, and whether all, some or none of the members of the class actually exist. 'reasonably well defined' is a difficult notion to pin down. I shall say only two things about it. Firstly that for a class of hypothetical entities to be 'reasonably well defined' it is not necessary that there be an effective procedure for determining conclusively whether or not any given hypothetical entity is a member. Secondly, that if there is nothing whatever which would count for or against the assertion that a certain hypothetical entity is a member of a given class, then that class is not 'reasonably well defined'. The procedure, however vague and woolly, whereby we try to establish whether or not an entity is a member of a given class I will call the 'membership criteria' of that class. It should be noted that this question is distinct from the question whether or not the class itself exists, and is also distinct from the question whether or not the class has any members.

4. The logical relationship between existence questions can now be stated more clearly. If the answer to the existence question for a certain class is that none of its members actually exist, then the existence question for any member or any subclass of that class must be answered likewise. If the answer to the existence question for a certain class if that all its members exist, then again its subclasses have the same property. If we know only that some members of a given class exist then the existence question for subclasses of that class remains open. It is important to appreciate that the relationship between classes which concerns us here is that of one class being a subclass of another class, not that of one class being a member of another. If a class is a member of another class then the existence question of the containing class is relevant to the question of whether or not the contained class itself exists, but it is not (or at least as far as I can see) relevant to what I have called 'the existence question' of the member class.

5. A consequence of this analysis, if it is correct, is that an adequate procedure for establishing the existence of an individual is to give grounds for an affirmative answer to the existence question of some class, and to show that the individual satisfies the membership criteria for that class. Now the criteria of existence which we use in our everyday dealings with the world are in fact membership criteria for various classes of entities. An affirmative answer to the existence question for these classes is assumed. I am assured of the existence of this typewriter because the fact that I can see and feel it is strong (if not conclusive) evidence that it is a member of the class of material objects, and to doubt the existence of material objects is an activity confined to philosophers.

6. Unfortunately membership criteria don't quite do all that we might hope for. They enable us to exchange one existence problem (say, the existence of my typewriter) for a more general existence problem (the existence of material objects), but they never give us an unequivocal solution. But membership criteria are all that we possibly can have, any criteria of existence does no more than define a class of entities which exist, i.e. it is the membership criteria of that class. We are always left with no demonstration that the members of the class actually do exist. Our ultimate criteria of existence must always stand defenceless, like the axioms of a logical system, plausible perhaps, but never demonstrable.

7. Though we cannot demonstrate the validity of our most basic criteria of existence, we can evaluate them on their consequences. Many mathematicians (myself included) do not require that the axioms of a mathematical system be 'self evident', or even plausible. It is sufficient if they produce satisfactory results. This is because a mathematician formalising a system already has a large number of results in mind which he wants to prove. A philosopher investigating ontological criteria (or for that matter, almost any philosophical problem), has his anchor in pre-existing linguistic conventions, and in the large body of generally accepted truths. We can judge his basic ontological doctrines at least partially by how well they fit into language as it is, and how well they fit with less contentious ontological issues.

8. This brings us to the criteria of ontological commitment. If we are to assess how well any given ontology agrees with ordinary discourse we must first settle the question 'what sort of beliefs constitute or entail ontological commitment to an entity or class of entities?'. The answer to this latter question must in turn depend upon what we hold to be the consequences of ontological commitment. To exemplify these relationships let us consider, first of all, one of Russell's doctrines. Russell thought that it was desirable to reduce to an absolute minimum the number of entities to which we are ontologically committed. He said, that wherever possible we should avoid ontological commitment to an entity by interpreting that entity as a logical construction from more primitive entities to which we are already committed. For example, material objects were to be logical constructions from sense data, and classes were to be eliminated in favour of propositional functions. the details of these reductions is not important for my present purposes, what is important is the intimate relationship between Russell's criteria of existence and his criteria of ontological commitment. If Russell wants to refrain from condemning ordinary usage concerning things such as tables and chairs then he must subscribe to a theory of ontological commitment which absolves ordinary discourse from ontological significance. To maintain such a spartan ontology as that of Russell it is probably best to maintain that ontological commitment is something which only ever happens which we are philosophising. That ordinary discourse neither presumes nor denies the reality of the entities it deals with. Under such a scheme of things the only way to commit oneself, ontologically, would be to assume the role of philosopher and explicitly pronounce on what does and does not exist. Such pronouncements made under ordinary circumstances would not count.

9. Another example of the relationship between ontology and ontological commitment stems from the belief that we cannot meaningfully talk about anything which does not exist. This belief imposes a rigid relationship between ontology, criteria of ontological commitment, and the meaningfulness of statements. Whenever we speak we either utter nonsense, or we commit ourselves to the existence of those things to which we refer. This is likely to lead us to a belief in all sorts of strange entities. The relationship between ontology and the consequences of ontological commitment may help us to dispense with such 'inflated' ontologies. This is done by asking odd questions about the entities which such an ontology postulates. When we have a sufficient number of unanswerable questions about these things our opponent might, with luck, lose faith in them himself. If however our opponent has properly looked after the relationship between his ontology, and what he holds to be the consequences of ontological commitment, then he will be immune to such attacks. The 'consequences' of importance here concern the legitimacy of asking various questions about allegedly existent entities. Then it would be reasonable for me to hold that a necessary consequence of the existence of some entity is that it makes sense to ask of that entity "What is its mass?". Conversely, if I accept that as a consequence of ontological commitment it becomes difficult for me to maintain an ontology which contains anything other than material objects. So the relationship is roughly this; the more we include in our ontology, the less we can say about the things which satisfy it. The most inflated of ontologies can survive if we hold that ontological commitment has no consequences.

10. The point of these illustrations is to suggest that even if we accept that our ontology should square up reasonably well with existing linguistic conventions, at least to the extent that we are not involved in denying the ordinary man's beliefs or in condemning his manner of expressing himself, we do not succeed in reducing ourselves to one possible ontological scheme. By manipulating the criteria for ontological commitment and the consequences of ontological commitment we can succeed in making a wide range of ontologies compatible with ordinary language and ordinary beliefs.

11. The main problem in this procedure is that 'manipulation' of the criteria for ontological commitment may involve us in a dual standard for the use of the verbs 'to exist' and 'to be', an ordinary standard, and a philosophical standard. Given that we have accepted the need to tailor our doctrines to fit in with ordinary language, why is it that these two verbs escape the restraint? Why should we be concerned to do any more than simply describe the ways in which these words are normally used? This takes us right back to the genesis of the philosophical dispute. Presumably the need for any philosophical corrective to the common usage of the verbs 'to exist' and 'to be' rests upon some deficiency or inconsistency in that common usage, and if this is the case, we could reasonably be expected to adopt that ontology which corrected these problems with the least possible alteration to common usage. What I am suggesting then is this; we should be content with the ontology implicit in ordinary use of the verbs 'to exist' and 'to be' unless we find this to be inconsistent. If we do find it inconsistent then there is some point in considering what would be necessary to render it consistent. I can see little value in suggesting alterations far more substantial than those necessary to render the use consistent. If anything common usage ontological criteria are likely to be much looser than their suggested counterparts in philosophical usage. At face value common usage commits itself to the existence of such things as possibilities, colours, differences, controversies, and indeed, anything at all that we might want to talk about. Now so long as we don't attach a great deal of significance to such existence few problems arise. If a philosopher wants to attach a certain significance to the existence of an entity then rather than restricting the range of existents he would do better to attach his significance to some subclass of the existents (e.g. material existents). If he wants to say something about everything that exists, then rather than trying to restrain our ontological commitments, he should restrain his ambitions and confine himself to some reasonably coherent subclass of existents.

12. The difficulties, if they are difficulties, which arise from the ordinary use of existential verbs seem to lie in the predicative use of the verb 'to be'. It seems necessary because of this to assert the existence of anything at all that we want to talk about (almost). One way of dealing with this is to separate the predicative use of the verb from its bald existential use, and confine ontological commitment to the latter. This solution may not necessarily suffice. I myself am strongly in favour of the abandonment of all consequences of ontological commitment. the question expressed when someone asks whether or not a given entity or class of entities exists can always be understood as a class membership question, where in many cases the class of entities concerned may be determined implicitly by the context. If the class whose membership we are disputing is not defined either explicitly or implicitly then the question lacks meaning. (Speaking qua philosopher in the context of some already enunciated ontological criterion is O.K. since the class whose membership is under dispute is the class of entities which satisfy the ontological criterion. Speaking qua philosopher in a debate about ontological criterion has to be understood as an attempt to set the context, a search for some specially important class of entities to be called 'those things which really exist' whose membership we can dispute ad. inf. I can see no unfortunate consequences flowing from the abandonment of ontological consequences. Even the round square is an incoherent concept, and we might well ask sensible questions about the membership of the class of incoherent concepts, (e.g. whether it is a subclass of the class of meaningless concepts or not, which is equivalent to the question whether or not there exists something which is both incoherent and meaningful, to which I am tempted to answer 'yes') Anything which we might be tempted to count as a consequence of ontological commitment we would do better to count as a consequence of membership of a certain class of entities. Even an ontology as drastic as that of Russell can readily be effectively absorbed into my anarchist ontology. By effectively absorbed I mean that we can abandon the ontology itself without loosing any of the discriminative power it represents, that we can obtain pragmatically equivalent results within a broader ontology. In Russell's case this would be done by discussing 'irreducible' entities without committing ourselves to the non-existence of 'reducible' entities. Such consequences as Russell attaches to the ultimate reality of an entity we would then attach to membership of the class of irreducible entities.

up home © RBJ written late 1974, edited into HTML 1997/4/6 last edited 2009/7/31