Essentialist Plenitude and the Semantics of Proper Names

A number of prominent metaphysicians have recently defended a set of ideas which I will call ‘essentialist plenitude.’ Very roughly, and to a first approximation, essentialist plenitude says that wherever there is an object with properties P 1 , … , P n there is in fact a plenitude of coincident objects that differ only in the distribution of essentiality and accidentality across P 1 , … , P n (§1). The main purpose of this paper is to arouse the suspicion that essentialist plenitude may have far-reaching consequences for the semantics of proper names. More specifically, I will argue that neither descriptivist nor causal theories of proper names work in a plenitude setting (§2). I will close by suggesting that our use of proper names may be thoroughly infected by semantic indecision about which among many coincident objects is referred to (§3).


Essentialism and Plenitude
A distinction is sometimes drawn between essential and accidental properties. Arguably, however, the more fundamental distinction is not between two kinds of property, but between two ways of instantiating a property (McGinn 2000 Ch. 4). In terms of the traditional image of instantiation as a 'metaphysical glue' binding objects and properties, the idea is that there are in fact two such glues: a low-grade a-glue and a high-grade e-glue. When an object and a property are a-glued, they can also be unglued, such that the object persists without the property; when they are e-glued, they can never be unglued. We might say: in the first instance we distinguish not between being essentially-F and being accidentally-F, but between being-essentially F and being-accidentally F.
As I will understand it here, essentialism is the thesis that there is indeed this specially robust way for an object to have a property, a way that does not permit undoing. One advantage of essentialism is that it can explain the apparent possibility of distinct but coincident objects. Call the collection of metal and other materials making up the Eiffel Tower 'Matt.' According to ' one-thingists,' Matt and the Eiffel Tower are one and the same object. Most philosophers are 'two-thingists,' though, holding that Matt and the Eiffel Tower are numerically distinct objects that happen to be collocated. The main reason for two-thingism's popularity is that Matt and the Eiffel Tower appear to have different identity and persistence conditions: if Matt were flattened into a wide disc, the Eiffel Tower would be no more, but Matt would persist. Essentialism has a nice explanation of this: although Matt and the Eiffel Tower (henceforth, ET) have so many properties in common, some of them they have in different ways. Thus, although both are S-shaped, Matt is-accidentally S-shaped whereas ET is-essentially S-shaped.
However, this phenomenon produces a conundrum for the essentialist. Suppose, as per two-thingism, that Matt and ET are distinct objects, and that what makes them distinct is that there are properties F and G such that Matt is-essentially F and is-accidentally G whereas ET is-accidentally F and is-essentially G. Question: Is there also a third object, coincident with both Matt and ET, which has both F and G essentially, and/or a fourth one that has both F and G accidentally? If not, why not, and if yes, where does this kind of ontological proliferation stop? The essentialist faces a dilemma: either (a) provide a principle of essence-restriction, whereby some distributions of essentiality and accidentality across any collection of co-instantiated properties are admissible while others are inadmissible, or (b) openly adopt a principle of plenitude, whereby virtually every logically coherent distribution of essentiality and accidentality corresponds to a distinct object. (Why 'virtually'? We will see momentarily.) A surprising number of prominent metaphysicians have recently opted for (b), embracing what I will call ' essentialist plenitude.' The basic idea is clearly stated by Karen Bennett (2004: 354): … every region of space-time that contains an object at all contains a distinct object for every possible way of distributing ' essential' and ' accidental' over the non-sortalish properties actually instantiated there. 1 The idea is further developed and defended by Sarah-Jane Leslie (2011) andMaegan Fairchild (2019), and is also endorsed by Kit Fine (1982: 100, 1999, Stephen Yablo (1987: 302, 310), Mark Johnston (2006: 697-8), and Shamik Dasgupta (2018: 547-8).
Leslie illustrates and motivates plenitude with a 100-plank ship of Theseus, each plank of which is replaced by another and immediately burned. What number of planks have to be replaced for the original ship to go out of existence? We avoid these familiar difficulties if we admit 100 coincident objects where the ship of Theseus is: one that would go out of existence if a single plank were replaced, one that would go out of existence if two planks were replaced, and so on.
Strictly speaking, plenitude allows many more than 100 objects here. For it admits not just one object that goes out of existence when two planks have been replaced, but many: that which goes out of existence when planks 1 and 2 have been replaced, one which goes out of existence when planks 2 and 3 have been replaced, and so on. Real plenitude thus appears to admit 2 100 coincident Theseus-ships. A general formulation of essentialist plenitude (henceforth, EP) might therefore be this: (EP 1 ) For any region R occupied by object O with properties P 1 , …, P n , there are in R exactly 2 n coincident objects, each with its unique distribution of essentiality and accidentality across P 1 , …, P n .
Unfortunately, this formulation cannot be quite right. For among an object's properties are many that depend on the object's identity. Modal properties are a case in point: because Matt is only accidentally S-shaped, Matt has the modal property of being possibly-flat; ET has the modal property of being necessarilynonflat precisely because it is essentially S-shaped. Accordingly, no object could coincide with ET that would be accidentally necessarily non-flat (or for that matter accidentally possibly-flat). There may be other examples (e.g., sortal properties, certain second-order properties). One way to solve this problem is to explicitly exclude these kinds of properties, which Bennett calls 'sortalish': (EP 2 ) For any region R occupied by object O with non-sortalish properties P 1 , …, P n , there are in R exactly 2 n coincident objects, each with its unique distribution of essentiality and accidentality across P 1 , …, P n .
But EP 2 faces a different difficulty: certain metaphysical principles might rule out otherwise possible distributions of essentiality and accidentality. A plausible principle rules out objects that have all their properties merely accidentally. Other metaphysical principles might rule out further logically coherent essential/ accidental distributions. Thus, both Bennett (2004: 357-8) and Leslie (2011: 279) seem to suggest that if an object has a determinate property F essentially, it cannot have any relevant determinable property merely accidentally. It is of course hard to know just how many candidate distributions will be ruled out without a complete list of essence-restricting metaphysical principles. Maegan Fairchild (2019) has recently offered the most developed attempt to formulate EP with precision. But I think the basic idea of essentialist plenitudeindeed the idea that Fairchild and others are trying to capture -is the following: when we compare the number of metaphysically permissible distributions of essentiality/accidentality to the number of metaphysically impermissible distributions, we find that the former vastly outnumber the latter. We might put this as follows: (EP 3 ) For any region R occupied by object O with non-sortalish properties P 1 , …, P n and, such that there are m logically permissible but metaphysically impermissible distributions of essentiality and accidentality across P 1 , …, P n , (i) there are in R exactly 2 n -m coincident objects, each with its unique metaphysically permissible distribution of essentiality and accidentality across P 1 , …, P n , and (ii) 2 n vastly outnumbers m.
The term 'vastly' is intentionally vague. The thought is that if m is so big that 2 n -m equals 2, or even 20, then not much is left of the idea of plentiful coincidence. But as long as m is dwarfed by 2 n , we have genuine plenitude (or 'near-plenitude,' if we prefer -I will ignore this terminological nuance).
Although it is not my purpose to defend EP, but rather to explore its implications for the philosophy of language, it might be useful to review briefly the motivation standardly adduced in its favor. Crushing the subtleties, Bennett's (2004: 354-9) basic argument is that there are only two stable positions on the problem of material constitution, one-thingism and all-out plenitude. 2 Leslie (2011: 278-80) argues, in a similar vein, that any substantial restriction of the distribution of essentiality and accidentality across a collection of co-instantiated properties would be arbitrary, and Johnston (2006: 697-8) remarks that 'the only nongerrymandered position in the domain is one that embraces a full plurality of principles of unity, which allows that the holding of any relation or property of some item makes for a whole.' The basic reasoning, then, is this: 1) The only non-arbitrary positions on material constitution are one-thingism and EP 3 ; 2) One-thingism has massively unacceptable consequences (see, e.g., Fine 2003); 3) EP 3 is to be preferred.
In this representation, Premise 1 may be seen as presenting a challenge to opponents of one-thingism who wish to avoid plenitude: Please provide a non-arbitrary principle that would rule in a relatively small number of coincidents.
Ernest Sosa (1999) may be read as addressing this challenge. Let a 'snowdiscall' be anything which is either a snowball or a snow-disc. Imagine now that some snowball in location L is flattened into a disc. The flattening destroys the snowball in L, but does not destroy the snowdiscall in L. Accordingly, the snowball and snowdiscall are numerically distinct objects, even before the flattening. Obviously, we can repeat this little procedure with any number of other introduced terms ('snowpyramidall,' 'snowcubeall,' etc.), leading to the inclusion in our ontology of innumerably many coincident objects in L. To avoid this unseemly ' ontological explosion,' Sosa proposes that we embrace ' existential relativity': an object's existence is relative to a conceptual scheme. Our conceptual scheme includes the concept snowball, but no snowdiscall concept, so relative to it there are snowballs but no snowdiscalls. Relative to other conceptual schemes, though, snowdiscalls may exist too.
Existential relativity would, if accepted, provide a principled restriction of essentiality/accidentality distributions across collections of properties. Let O j be a putative object which has P 1 , …, P j essentially and all its other properties accidentally. According to Sosa's principle, O j exists relative to conceptual scheme S if, and only if, S includes a concept C, such that C's application conditions 'match' O j 's identity conditions (read: C's application conditions are exactly the conditions fulfillment of which guarantees O j 's existence). Saying 'O j exists' simpliciter might then be analyzed as elliptical for 'O j exists relative to my/our conceptual scheme' or some such. Now, the obvious problem here is that it is somewhat mysterious how a concept can make a material object exist which would be nonexistent had there been no such concept. It is much more plausible to think of conceptual schemes as selecting objects for inclusion in our 'folk ontology.' Perhaps once there is a plenitude of objects, we rely on our concepts to select the subset most worth keeping track of. But the objects from which the selection is made must be there before our concepts come to them. In other words, existential relativity applies more plausibly to so-called ordinary objects: for something to qualify as an ordinary object (i.e., to be an object in our folk ontology), there needs to be a matching concept in our conceptual scheme. Still, the conceptual scheme cannot make it the case that this 'something' is there in the first place. We may say that it is the ordinariness of an ordinary object, not its objecthood, that is conceptual-scheme-relative. 3 Sosa had a reason, of course, for adopting existential relativity: we ought to avoid ' ontological explosion' -which is just a less sympathetic name for 'plenitude'! The proponent of plenitude might respond, however, that parsimony only requires us not to multiply entities without necessity. If, as the plenitudinist claims, one-thingism has unacceptable consequences and 'few-thingism' is unstable, then there is theoretical necessity to embrace plenitude.
There is no question, of course, that EP is a highly surprising, unintuitive thesis. But that does not exactly separate it from many other theses in contemporary metaphysics (think of mereological nihilism and universalism, priority and existence monism, ontic structural realism, etc.). For the remainder of this paper, I propose that we indulge EP and consider some of the consequences it might have for the philosophy of language. This paper's thesis is but conditional: if EP is true, then…

Essentialist Plenitude, Descriptivism, and Direct Reference
In this section, I argue that if EP is true, then neither descriptivist nor causal approaches to nominal reference are viable as they stand. In itself, this conditional contains no concrete recommendation as to whether we should reject EP or revise our best theories of reference. I leave it to the reader's antecedent commitments to determine under which light to see the import of the argument. My goal in this section is thus merely to point out a certain reflective disequilibrium.

Descriptivism and essentialist Plenitude
If you show the average person a picture of ET and ask what the picture represents, you will probably be answered 'The Eiffel Tower.' If you then ask 'How do you know?,' you are liable to hear something like 'It looks like the Eiffel Tower.' Even if this answer is philosophically naïve, a picture's natural resemblance to its referent is surely useful to its representational ability.
In contrast, the name 'the Eiffel Tower' bears no resemblance to ET. How does it nonetheless manage to refer to it? Descriptivism answers roughly as follows. Every proper name has not only a reference, but also a descriptive content, which plays double duty for the speaker. First, it is grasped by the speaker as she uses the name, thereby capturing the name's cognitive significance to the speaker. Secondly, it specifies a condition any object must satisfy in order to qualify as the name's referent. Typically, the thought goes, exactly one object will satisfy the description (even if nothing guarantees this). This is how we manage to use proper names to refer to worldly objects.
We may formulate descriptivism more precisely as the conjunction of three theses.
(1) For any name N, there is some definite description D, such that the semantic value of N is identical to the semantic value of D.
The 'semantic value' of a name N is whatever N contributes to the proposition expressed by assertions featuring N. According to (1), 'the Eiffel Tower' has the same semantic value as some definite description D. Consequently, 'The Eiffel Tower is in Paris' and 'D is in Paris' express the same proposition.
(2) The semantic value of a definite description D (a) is not identical to D's referent but (b) typically determines D's referent.
Although a definite description typically manages to select a single object, it is not this object that it contributes to the proposition it expresses, but some other entity -a sense, mode of presentation, intension, or some such.
(3) The semantic value of a definite description D is typically graspable by the speaker who utters D.
That is, competent use of a description typically involves the user's being in a position to understand the condition specified by the description. For our purposes, any theory of reference that respects (1)-(3) is a form of descriptivism. Kripke (1972) famously presented a formidable case against descriptivism. We will discuss his arguments toward the end of this subsection. What I want to argue first is that if EP (essentialist plenitude) is true, then there is very little chance that (1)-(3) could be co-satisfied. The basic problem is this: Plenitude produces such a vast number of potential description-satisfiers, that selecting a unique referent requires highly intricate descriptions -so intricate that they become hardly graspable by ordinary speakers.
The point is rather that doing so should suffice to do justice to whatever insight one might recognize in Sosa's reasoning.
Consider first what is probably the most naïve version of descriptivism, construing a name's descriptive content as one central fact about the name-bearer. For instance, 'the Eiffel Tower' might refer to whatever satisfies 'the tower that Eiffel built.' In this form, (2) is certainly not met under the assumption of plenitude. For if EP is true, then in the relevant region there are incredibly many objects with the property of having been built by Eiffel. We could make some progress by relying on a more sophisticated description such as 'the x, such that x is essentially a tower that Eiffel Built.' But in a plenitude setting there are still too many objects answering to this description. There is the tower that is essentially built by Eiffel and is essentially in Paris, but also the tower that is essentially built by Eiffel and is only accidentally in Paris. (If in 2050, post-Brexit England invades France, conquers Paris, and moves the Eiffel Tower to Trafalgar Square, the first object would go out of existence but the second would persist.) The only way to secure uniqueness is to appeal to descriptions rich and explicit enough to specify a complete distribution of essentiality and accidentality -something of the form 'the x, such that for any property P that x has, if P = being built by Eiffel, then x is-essentially P, and if P ≠ being built by Eiffel, then x is-accidentally P.' A version of descriptivism that identified this as the descriptive content of 'the Eiffel Tower' would satisfy (2). The problem is that my four-year-old is a competent user of the name 'the Eiffel Tower,' but cannot grasp the description just mentioned, nor any colloquial/imprecise degeneration thereof (e.g., 'the tower essentially built by Eiffel and accidentally anything else'). In this way, the reference-determination and grasping conditions produce pressures in opposite directions, making it hard to envisage their co-satisfaction in a plenitude setting. Call this the problem of graspable reference-determination.
A more sophisticated version of descriptivism appeals to cluster-descriptions built out of salient facts about the name-bearer (Searle 1958). Suppose for simplicity that ET has only five salient features: (a) it is in Paris, (b) it is S-shaped, (c) many people visit it, (d) it is tall, and (e) it is called 'the Eiffel Tower.' Importantly, the ' cluster' of (a)-(e) is not the simple conjunction of them, but something like the disjunction of all conjunctions of most of them. So in a simple version, the descriptive content of 'the Eiffel Tower' would be 'the thing which is either (a), (b), and (c); or (a), (b), and (d); or (a), (b), and (e); or (a), (c), and (d) …' -fill in all sixteen possible combinations of having three, four, or five among (a)-(e). Given EP, however, there are ever so many coincidents that have at least three among (a)-(e). Here too we could reframe the description in explicitly essentialist terms: 'the thing which is either essentially (a), (b), and (c) and accidentally anything else; or essentially (a), (b), and (d) and accidentally anything else; or essentially (a), (b), and (e) and accidentally anything else …' But this faces two problems. First, my son still cannot grasp this kind of description. And second, if EP is true, then corresponding to the sixteen ways of having at least three among (a)-(e) essentially and any other property accidentally are sixteen different coincident objects. So the relevant description still fails to uniquely determine reference.
I conclude that proponents of EP should reject standard versions of descriptivism. But do they need this EP-based argument against descriptivism, given Kripke's decisive case against it? I think they do. Descriptivist theories have actually proved remarkably resilient in the face of Kripke's arguments. His main ('modal') argument was that the difference between names and descriptions becomes evident in modal contexts: 'The tower that Eiffel built might not have been the Eiffel Tower' expresses a true proposition, whereas 'The Eiffel Tower might not have been the Eiffel Tower' expresses a false proposition (Kripke 1972: 48-9). Many descriptivists have argued, however, that this only forces them to rigidify their descriptions (Jackson 1998): instead of 'the tower that Eiffel built,' it would be 'the tower that Eiffel actually built.' For 'The tower that Eiffel actually built might not have been the Eiffel Tower' is false as well. But note, now, that rigidification is of no help in the face of EP: the many coincidents in the relevant region satisfy the rigidified description just as much as the non-rigidified one (they are all actual!). Kripke's (1972: 78, 87) second main argument was epistemic: 'The Eiffel Tower is the Eiffel Tower' is a priori; not so 'The Eiffel Tower was built by Eiffel.' However, as Searle (1983 Ch.9) noted, this argument fails against cluster descriptivism. For a sufficiently large, sufficiently textured cluster of properties saliently attributed to ET, it is much more plausible to consider it a priori that ET has sufficiently many of them (see already Searle 1958: 168). As we have seen, however, if EP is true, then cluster descriptivism is undermined by its failure to deliver reference-determination. I conclude that EP represents an acute challenge to descriptivist theories of reference, a challenge which goes beyond Kripke's critique and infects equally post-Kripkean versions that appeal to rigidified and/or cluster descriptions. 4

Direct Reference and essentialist Plenitude
The pars destruens of Kripke's Naming and Necessity involved the claim that proper names refer without the mediation of any description, hence ' directly.' What this means is that the semantic value of a proper name is exhausted by its referent. This raises the question, though, of how the referent of a token of a name might be selected if not by the satisfaction of a condition specified by its descriptive content (and not, obviously, by the name's resemblance to its referent). The pars construens of Naming and Necessity was an answer to this question: roughly, there is a specific causal connection that links a name with its bearer. The speaker need not be aware of this causal connection, yet it is this connection that determines the referent for the name she uses.
To get a clear sense of the relevant causal connection, it is useful to divide the Kripkean story into two parts: an account of reference-fixing and an account of reference-borrowing. Reference-fixing occurs when an object is originally assigned its name in a baptismal event. Alexandre and Catherine-Mélaine Eiffel might, for instance, have ostended their new baby and announced 'Thou shall be named Gustave Eiffel.' Referenceborrowing happens when, through a series of causal exchanges, the Eiffels' friends and neighbors start calling the baby Gustave, then his teachers and peers do, and so on all the way to our present-day uses. The exact mechanics of this is something Kripke is programmatic about, but the basic idea is clear: a name's referent is not determined by any description, but by a complex causal chain linking use of the name to the referent itself.
This causal theory faces well-known difficulties in producing correct predictions in myriad cases (see, e.g., Evans 1973 on 'Madagascar'). The Eiffel Tower was apparently originally named 'the 300-meter tower' in all official documents and ceremonies. All the same, right from the start people seemed, quite spontaneously, to prefer referring to it as 'the Eiffel Tower' (La Tour Eiffel). Thus our current use of the name is not causally chained to any orderly baptismal event, but derives its reference from a diffuse field of popular tendencies in late 19 th -century Paris.
Still, these kinds of difficulty have struck proponents of the causal theory as amenable to technical resolution (Evans 1973;Devitt 1981). What I want to argue now is that essentialist plenitude produces a much more principled difficulty for the causal approach.
Suppose, for the sake of simplicity, that a solemn, orderly ceremony took place in 1889 in which the French president pointed at the Eiffel Tower and announced 'Thou shall be named the Eiffel Tower.' If EP holds, the lackluster president was in fact pointing at a gazillion different coincident objects. There is nothing about his act to select one among them as the rightful name-bearer. Thus even if the appropriate causal chain did link my current use of 'the Eiffel Tower' to such a baptismal event, my use of the name would still not refer to any unique object. Call this the problem of mass baptism.
Obviously, Kripke's baptism story is just one version of the causal theory. But other versions fare no better. Devitt (1981), for instance, claims that the causal chain is continuously anchored in acts of perceptual encounter, without a single privileged baptismal event. Still, every perceptual encounter with the Eiffel Tower is an encounter with a great many coincidents. So the problem recurs.
In the remainder of this section, I consider three possible responses to the problem of mass baptism: two that try to identify a single object among the many coincidents as the rightful name-bearer and one that tries to reconceptualize proper names in a plenitude setting as somehow involving plural reference.
Response 1: The essentiality of origins. Kripke (1972 fn 57) held that material objects such as ET have essentially (i) their material origins (e.g., originating from the relevant quantity of iron), (ii) their 'material substantial constitution' (e.g., being iron), and (iii) their kind (e.g., a tower) -and have accidentally all other properties. Kripke might therefore suggest that the Eiffel Tower is whichever baptized coincident had essentially these three properties and accidentally all its other properties.
This approach faces, however, the kind of ship-of-Theseus dynamic belabored by Leslie (2011), and consequently returns the wrong results in certain scenarios. Suppose that after ET's inauguration, Monsieur Dupont surreptitiously removes one strip of iron from the tower each month, replacing it with an indistinguishable strip and keeping the original on a private island in the Indian Ocean. Toward the end of his improbably long life, and now in possession of all the iron from the original tower, Dupont secretly builds on his island an enormous structure in the shape of the existential quantifier. The view under consideration would return the result that ET has migrated from Paris to the Indian Ocean and is now shaped like a backward E. Intuitively, however, ET is still in Paris and if anything is shaped like an upside-down universal quantifier. reference face considerable difficulties if EP is true, difficulties which will be inherited by causal descriptivism.
Response 2: Naturalness and reference-magnetism. David Lewis (1983) included in his ontology a plenitude of properties. A property, for Lewis, is a set of individuals, and every set of individuals, however gerrymandered, constitutes a property. To avoid problems of indeterminacy of predicates' reference to properties, he then introduced a primitive notion of naturalness, such that natural properties magnetize predicates' reference (Lewis 1983(Lewis , 1984. A causal theorist of reference could suggest that of all the coincidents in the relevant Parisian region, one is the most natural and therefore bears the name 'the Eiffel Tower' exclusively. (It is unclear that the result is still a causal theory, but let us bracket that issue.) There is an important disanalogy, however, between the case of Lewisian properties and our case of plenitudinous objects. With Lewis' properties, we have an intuitive grip on naturalness, in the form of objective resemblance: if S is the set of all and only square objects and S* the set of all square objects plus the moon, we understand why S is more natural than S*. In contrast, if ET 1 has P 1 , … P j -1 essentially and P j , …, P n accidentally; ET 2 has P 1 , … P j essentially and P j + 1 , …, P n accidentally; and ET 3 has P 1 , … P j + 1 essentially and P j + 2 , …, P n accidentally, we have no sense of what it would mean to claim that one of them is more natural than the others. There is no analog of objective resemblance here -nor any other alleged symptom of naturalness -to help us. (Might there still be some naturalness inherent in the objects themselves, though entirely opaque to us? This can always be said, of course, but it is hard to see what reason we might have to believe that it is indeed the case.) Response 3: Plural reference. The causal theorist might simply suggest that reference could be causally grounded in a mass baptism in which innumerably many coincidents are co-named. Instead of the solemn, intimate event envisaged in 'Thou shall be named the Eiffel Tower,' we would have the kind of lowbrow affair stamped with 'All y'all are from now on the Eiffel Tower.' If reference-borrowing functions as it should, current uses of 'the Eiffel Tower' would then refer at once to each among the many relevant coincidents. This kind of plural reference has been offered as accounting for the semantics of other kinds of expression, for example mass nouns. A mass noun such as 'water,' it has been claimed (Laycock 2006), refers plurally to all quantities of water. The causal theorist of reference might allow proper names to function the same way: 'the Eiffel Tower' could refer plurally to all those coincident towers.
The obvious problem here is the lack of any psycholinguistic realism. Proper names are plausibly the paradigms of singular terms. Unsurprisingly, then, they fail the standard grammatical tests for plural expressions. The most obvious test is that plural expressions cannot be pluralized, being already plural. Consider 'the apples' (as in 'The apples are delicious'). You cannot say 'the appleses' -such double pluralization is grammatically impermissible. Proponents of plural-reference accounts of mass terms, for instance, have pounced on this phenomenon: while 'The water is warm' makes sense, 'The waters are warm' seems infelicitous. 5 In contrast, however, proper names do not resist pluralization: 'The Obamas ran a scandal-free White House' is perfectly felicitous. This is because proper names are clearly singular rather than plural terms.
(Might 'Eiffel Tower' refer to the single collection, or group, of all the relevant coincidents? If ' collection' or 'group' means set, then this is unlikely, since sets are abstract objects whereas 'Eiffel Tower' is supposed to be the name of a material object. If they mean rather sum, it is hard to think about the view clearly, since it is unclear exactly what the sum of a and b is when a and b are distinct but have all the same parts.) I conclude that EP and causal theories of reference are uncomfortable bedfellows. As we saw, this is also true of EP and descriptive theories of reference. Should we simply reject EP to solve the problem? That would be strange. EP is an ontological theory about what there is in the world. On the face of it, the question of what there is in the world is not beholden to considerations concerning what would raise fewer problems for our talking about the world.

Proper Names and Semantic Indecision
In this final section, I want to consider a more radical possibility: that if a proper name is a singular term, but there are many objects which (a) are suitably causally linked to it and (b) satisfy any graspable descriptions associated with it, then there must be some kind of semantic indecision involved in our use of the name. Thus, 'the Eiffel Tower' refers to one among any number of coincidents, without requiring us to waste cognitive resources on determining exactly which one.
This kind of semantic indecision is familiar from the vagueness literature. Lewis (1986: 213) appealed to it memorably in analyzing the reference of vague singular terms: The only intelligible account of vagueness locates it in our thought and language. The reason it's vague where the outback begins is not that there's this thing, the outback, with imprecise borders; rather there are many things, with different borders, and nobody has been fool enough to try to enforce a choice of one of them as the official referent of the word ' outback.' Vagueness is semantic indecision.
What concerned Lewis is the vast plurality of almost-coincident objects with equal claim to being the referent of 'the outback.' What concerns us here is the vast plurality of totally-coincident objects with equal claim to being the referent of 'the Eiffel Tower.' In both cases, natural language has not selected one among the claimants. Our semantic-indecision claim is stronger, however, as it applies to all proper names regardless of any vagueness issues. Even names whose bearers have precise borders, such as 'Australian Capital Territory,' involve semantic indecision. At bottom, it is indecision about which of the many properties instantiated in that territory are instantiated essentially and which merely accidentally.
From a pragmatic perspective, this is rather to be expected. Coincident objects share all their parts, as well as all their ('non-sortalish') properties. If you have seen one legitimate 'Eiffel Tower' claimant, you have seen them all. If you have scaled one, you have scaled them all. There is no need for you to encumber your idiolect with a distinct name for each 'Eiffel Tower' claimant.
It is worth distinguishing here two versions of the semantic-indecision view, which we may call 'universalist' and 'pluralist.' The universalist takes every coincident in the relevant region to have a legitimate claim to be the referent of 'the Eiffel Tower.' The pluralist allows 'the Eiffel Tower' sufficient decisiveness to rule out many coincidents, but not enough to rule in exactly one. The pluralist version will probably match intuitive verdicts better. For instance, the object which has all the Eiffel Tower's properties essentially (and none merely accidentally) does not seem like a legitimate claimant. For the structure has any number of properties the loss of which would nowise incline us to stop applying the name 'the Eiffel Tower.' This means that nothing which answers to 'the Eiffel Tower' would go out of existence in conditions in which the coincident that has all of ET's properties essentially would. So this coincident has identity and persistence conditions different from those of anything that may legitimately claim to bear the name 'the Eiffel Tower.' The issue here, it is important to stress, is not metaphysical. Metaphysically, we have agreed to consider virtually every distribution of essential/accidental as 'implemented' in an object. The question is linguistic: whether our proper names tolerate reference to all or only some of these 'implementers.' I suspect 'only some' is the correct answer. Still, unless in some special cases only one implementer is tolerated, our proper names will always involve semantic indecision as to which among the many tolerated implementers is referred to.
How should we understand the attribution of truth-values to statements featuring proper names, if the latter involve semantic indecision? There are a number of ways to go here, roughly corresponding to the different approaches in the parallel case of vague singular terms. The best strategy for the proponent of EP is probably to let that debate play itself out and plug the 'winner' into her view. My initial preference, though, would be for something like a supervaluationist story: 'The Eiffel Tower is in Paris' is true because every object with a legitimate claim to being the referent of 'the Eiffel Tower' is in Paris; 'The Eiffel Tower is in London' is false because false of all legitimate 'Eiffel Tower' claimants. In general, given that all relevant coincidents share their non-sortalish properties, predications of such properties will be true of all claimants when true of one of them (and false of all when false of one). For all we have said, however, other kinds of predication (e.g., modal) may result in 'truth-value gaps': it would be the case neither that all legitimate 'Eiffel Tower' claimants satisfy the predicate nor that all fail to satisfy it.
These are more speculative remarks about the possible consequences of essentialist plenitude for the semantics of proper names. The main aim of this paper, recall, has been to arouse the suspicion that a metaphysical thesis recently defended by a number of prominent metaphysicians may have far-reaching consequences for the semantics of proper names. I hope the above discussion manages to arouse this suspicion. I have argued that if essentialism indeed leads to plenitude, then both descriptivist and causal theories of reference will find it hard to generate unique referents for proper names, and pervasive semantic indecision may attend our (competent) use of proper names. This notion is a departure from common sense, of course, but if I am right, the only way to avoid it is to deny either (a) essentialism or (b) the line of reasoning leading from essentialism to plenitude or (c) the line of reasoning, aired herein, leading from essentialist plenitude to semantic indecision in proper names. 6