At Sophist 247e, Plato puts the following into the mouth of the Eleatic Stranger:
I suggest that anything has real being that is so constituted as to possess any sort of power either to affect anything else or to be affected, in however small a degree, by the most insignificant agent, though it be only once. I am proposing as a mark to distinguish real things that they are nothing but power. (Cornford tr.)
The gist of the passage is that what makes a thing real or existent is its (active) power to affect other things or its (passive) power to be affected by them. In sum,
D. For any x, x exists =df x is causally active or passive.
Thus everything causally active/passive exists, and only the causally active/passive exists. The definition rules out of existence all 'causally inert' items or abstracta such as propositions as Gottlob Frege construes them, namely, as the senses of context-free indicative sentences. And of course it rules out sets of Fregean propositions, and indeed all mathematical (as opposed to commonsense) sets. But what about the mathematical set of the books on my desk? Each of the books is existent or real by (D) and so is the object resulting from the physical bundling of the books together; but the mathematical set of these books is abstract and thus causally inert. So if (D) is true, we cannot admit mathematical sets into our ontology. For such items cannot enter into causal relations. Fregean propositions and mathematical sets are therefore putative counterexamples to (D). If these counterexamples are genuine, then (D) fails extensionally: the extension of the existent is wider than the extension of the causally active/passive.
But what interests me at the moment is not the question of the extensional correctness of (D) but a deeper question. Even if we assume that (D) is extensionally correct, i.e., that all and only existents are causally active/passive, does (D) tell us what it is for an item to exist? When we say of a thing that it exists, what are we saying about it? Are we saying that its existing consists in its being causally active/passive? My answer is in the negative — even if it is true that all and only existents are causally active/passive.
My reason is quite simple. For an item to be capable of acting or being acted upon it must 'be there' or exist! 'Before' it can be a doer or a done-to it must exist. (The 'before' is to be taken logically not temporally.) The nonexistent cannot act or be acted upon. There is no danger that winged horses will collide with airplanes. The reason is not that winged horses are abstract or causally inert objects; the reason is that they do not exist. Winged horses, if there were any, would belong to the category of the causally active/passive. Horses, winged or not, are animals, and animals are physical things located in spacetime. But winged horses don't exist — which is the reason why they cannot act or be acted upon. Existence, therefore, is a necessary condition of an item's being a causal agent or patient. It follows that existence cannot be explicated in terms of power as per the Eleatic Stranger's suggestion. Existence is too fundamental to be explicated in terms of power. Causal activity/passivity presupposes existence. This is true not only of an attempted explication of existence in terms of power, but of any attempted explication of existence in terms of anything else. I’ll give a second example shortly.
There is a tendency to conflate two different questions about existence. One question about existence concerns what exists. Answers to this question can be supplied in the form of definitions such as (D) above. What (D) tells us is that all and only the causally active/passive exists. But there is a deeper question about existence, namely, the question as to what it is for an existing thing to exist. What I have just argued is that this second question cannot be answered with any such definition as (D). For even if you find a definition that is extensionally correct and immune to counterexamples, you will at the very most have specified the necessary and sufficient conditions for a thing's being among existing things. At most you will have answered the question as to what exists. You will have not thereby have put your finger on what it is for an existing item to exist. At most you will have answered a question pertaining to the ontological inventory. You will not have answered the question as to what qualifies an item to be listed in the inventory. Now for my promised second example.
Suppose you say that,
For any concrete individual x, x exists =df x has properties.
Meinongian scruples aside, this proposal has an excellent chance of being extensionally correct: necessarily, everything concrete that exists has properties, and everything concrete that has properties exists. But the proposal, which I take to be extensionally correct, does not get at (succeed in explicating) the existence of an existing concrete thing precisely because it presupposes the existence of existing concrete things. Given that concrete things exist, the biconditional is true of them. But if nothing concrete existed, the biconditional would still (logically speaking) be true. This is because all such putative definitions of existence (i.e., of what it is to exist) are, despite appearances, really circular: to capture what it is for a concrete individual to exist the definition would have to have the form:
For any x, x exists =df x is ____ and x exists.
For if you left off the codicil ‘and x exists,’ then the most you would have accomplished to to specify the necessary and sufficient conditions of an item’s admission to the ontological inventory, but not that which qualifies it for admission to the inventory, namely, its existence. No matter what you put into the gap, the explication moves in an explanatory circle inasmuch as the explication must presuppose that the x in question is an existing x. Here is a third example :
For any concrete individual x, x exists =df x is spatiotemporally located.
Even if true, this biconditional does not capture what it is for a concrete individual to exist. For while there are no winged horses, there might have been, and being what they are, they would have to be spatiotemporally located. The existence of a spatiotemporal item cannot therefore consist in its being spatiotemporally located. Its being spatiotemporally located is at best a consequence of its existence. What it is for an existing item to exist thus eludes definitional grasp.
One final example. Some will be tempted to say that the existence of a thing = its self-identity. But this proposal too is a non-starter. For while it is certainly true that nothing could possibly exist that was not self-identical, surely the actual existence of a thing cannot consist in its self-identity. For if that were the case, then everything that actually exists would be a necessary being, which is presumably something we do not want to say. There is at most only one being, God, whose self-identity is identical to its existence. Everything other than God is such that its existence and its self-identity are really distinct. So, while every existent is self-identical, no existent, except God, is such that its existence can be non-circularly explicated in terms of self-identity.
For more on this fascinating topic, see my A Paradigm Theory of Existence (Kluwer 2002), pp. 2-8.
To sum up. The question as to what exists must not be confused with the question as to what it is for an existing item to exist. The Eleatic Stranger’s answer to the first question in terms of power may or may not be true. But whether or not it is true, it cannot give us an answer to the second question. And the same goes for every other attempt to give a non-circular explication of existence. Existence is just too basic to be conceptualized in non-existential terms.