SciELO - Scientific Electronic Library Online

vol.39 issue1How to (dis)solve Nagel's paradox about moral luck and responsibilityParadoxical versus modulated conditional inferences: An explanation from the Stoicism author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand




Related links



Print version ISSN 0100-6045On-line version ISSN 2317-630X

Manuscrito vol.39 no.1 Campinas Jan./Mar. 2016  Epub May 17, 2016 


Singular Reference Without Singular Thought*

Filipe Martone1 

1University of Campinas (UNICAMP). Department of Philosophy. Rua Cora Coralina 100, Campinas - SP. 13083-896, Brazil.


In this paper I challenge the widespread assumption that the conditions for singular reference are more or less the same as the conditions for singular thought. I claim that we refer singularly to things without thinking singularly about them more often than it is usually believed. I first argue that we should take the idea that singular thought is non-descriptive thought very seriously. If we do that, it seems that we cannot be so liberal about what counts as acquaintance; only perception (and memory) will do. I also briefly discuss and reject semantic instrumentalism. Finally, I argue that while singular reference is cheap, singular thought comes only at a price.

Keywords: reference; singular thought; acquaintance; epistemology; semantics

The current debate about singular thought is roughly divided between those who believe that there is some kind of restriction on singular thinking and those who do not. Typically, those who advocate some restriction agree that we can have singular thoughts not only through perception and memory but also through communication chains and (perhaps) through causal traces of an object as well. I will call this position the conservative view, given its status as the orthodox position 1. Those who reject any restriction on singular thinking argue that the mere manipulation of direct reference devices is enough to guarantee singular thought. This view is traditionally called semantic instrumentalism. Even with the radical differences between the two views, both of them seem to agree that the conditions of singular reference and singular thought are more or less the same: they believe that whenever I am in a position to make a singular reference I am also in a position to have singular thoughts about the referent. The same mechanisms that secure singular reference, they hold, also secure singular thought. Thus, the content of our thoughts mirrors the semantic content of language. I call this the mirroring intuition.

In this paper I argue that this intuition is mistaken. To do that, I show that in many cases singular reference is not a guarantee of singular thought. My view on singular thought, then, is much stricter than the one that is commonly accepted. Against conservatives and semantic instrumentalists, I hold that singular thought is possible only through perception and memory. My general strategy is to argue for the following conditional: If we accept the basic definition of singular thought as non-descriptive thought and what I call the Russellian Motivation, then many cases that are accepted by conservatives and semantic instrumentalists as cases of singular thinking are better seen as cases of descriptive thought, even though singular reference still occurs. This leads to a split between language and thought and a rejection of the mirroring intuition. Thus, to maintain their position, the conservatives and semantic instrumentalists must reject the basic definition (or try to amend it) and reject the Russellian Motivation, or argue that the split between language and thought I am proposing is far too problematic to be accepted. These moves will be discussed and dismissed.

1.What is Singular Thought?

First, we must find some definition of singular thought that can be accepted by conservatives and semantic instrumentalists alike. This is necessary because it is easy to find in the literature many definitions that are already theory-laden, such as: (1) singular thought is a thought that uses a mental file; (2) is an attitude towards a singular proposition2; (3) is an object-directed thought, etc. Of course, any of these definitions may end up being correct. However, they are already theoretical interpretations of the phenomenon we are interested in, so none of them is a good starting point for our investigation. We need something more basic and yet compatible with all those items above.

So what exactly is the phenomenon we are concerned with and which definition is adequate to capture it without being compromised with any theoretical standpoint? For certain purposes, sometimes is useful to define something negatively, that is, by giving a definition that simply states what the definiendum is not. This strategy obviously presupposes that what the definiendum is not is clearer or better understood than the definiendum itself. This is precisely the case here. Since singular thought is universally believed to be somehow opposed to descriptive thought, we can start by defining it simply as non-descriptive thought3. This negative definition works reasonably well because descriptive thought is arguably a clearer notion than singular thought. To give an example, I think descriptively when I think (without having anyone in particular in mind) "the ugliest man on Earth must be lonely"4. In this case, my thought is about whatever object happens to satisfy the condition expressed by the description "the ugliest man on Earth", i.e., the condition of being the unique object that is both male and uglier than every other male on our planet. Another famous example of descriptive thought is the thought "the shortest spy is a spy". In both cases my thought "lands on" or "reaches" the relevant object by selecting and entertaining certain concepts that the object uniquely exemplifies. To use Recanati"s phrase, I think of the objet qua instantiator of properties, or qua satisfier of concepts. In sum, descriptive thought is the type of thought that is about an object because it fits some description. The negative definition, therefore, merely states that singular thought is the type of thought that does not need descriptions to be about something. This seems plausible enough, and I believe most (if not all) philosophers would agree with it5.

Not surprisingly, this distinction echoes Russell"s famous distinction between knowledge by acquaintance and knowledge by description, which is the origin of our current debate about singular thought. If this is the case, then it is not unexpected that our modern way of drawing this distinction owes a great deal to Russell. To see how this is so, let us take a look at some of Russell"s concerns. I believe that an inspection of his original motivations can shed some light on why we started distinguishing between singular and descriptive thought in the first place.

Russell was among other things an epistemologist. As such, he was concerned with the ways by which we operate with our cognitive apparatus in order to think about things. He wanted to investigate the conditions under which it is possible to think about objects that are not only different in kind but also that bear different relations to our cognition. After all, the world is populated with all sorts of things, bearing all sorts of relations to our thought. Because we are finite and epistemically limited beings, it seems clear that we cannot think about all of them in the exact same manner. We are not omniscient. Some things seem easily accessible to us, such as sense-data, and some, more remote things, require more effort to be thought about. Some objects, for example, are accessible only via fairly sophisticated descriptions, such as "the center of mass of the Solar System" or "the smallest prime between 250³²² and 250³²³". So, some objects can be accessed more immediately and directly, while others require some conceptualization in order to be grasped. This is simply a consequence of our limited cognitive nature.

Russell had this in mind when he introduced the distinction between knowledge by acquaintance and knowledge by description. This distinction is meant to characterize two modes of thinking and of acquiring knowledge about the different objects in the universe. In the light of the preceding considerations, what knowledge by description is seems clear enough: it is the kind of knowledge we have about things that are more or less remote, which can be grasped only via the properties they exemplify. On the other hand, knowledge by acquaintance is the kind of knowledge we have by standing in some more immediate epistemic relation to things. For Russell, we have knowledge by acquaintance when we are "directly aware" of an object, when the object is "smacked" against our minds, so to speak. If we bear such a relation to something, we do not need to conceptualize it or think of it under any description. The object is simply "given" to our thought. Thus, knowledge by acquaintance is, essentially, non-descriptive knowledge. His notion of knowledge by acquaintance, then, is nothing but a version of singular thought; and a very restrictive one, because he accepted only sense-data and universals as possible objects of acquaintance. So, it is the distinction between knowledge by acquaintance and knowledge by description - which lies at the heart of Russell"s foundationalist epistemology - that laid the grounds for our contemporary distinction between singular and descriptive thought.

Notice that the distinction between singular and descriptive thought, as made by Russell, is not a matter of how we choose to think about objects. As said earlier, we are epistemically limited beings and the universe contains a remarkable variety of things. Given that, Russell is interested in how it is possible to think about those things and in how our knowledge of them comes about. In his view, we cannot simply choose to think about something descriptively or non-descriptively. This is made very clear in his "Bismarck" example. He says: "Here [when Bismarck makes a judgment about himself], the proper name ["Bismarck"] has the direct use which it always wishes to have, as simply standing for a certain object, and not for a description of the object" (Russell, 1949)6. This excerpt is very telling: Russell is claiming that it is only possible to have a singular thought about an object when one is acquainted with that object, as Bismarck is with himself. In any other cases - and this is the crucial idea - we always wish to have singular thoughts, but we are forced to think descriptively due to our limited cognitive capacities. If we were God, presumably, we would not need to think of anything under any description; all objects in the universe would simply be "given" to our thought. Our thinking, then, is constrained by the type of relation that the object bears to our minds and by its kind, insofar as the kinds objects fall into limit the sort of epistemic relation we can bear to them. This epistemic relation then determines how we should exercise our cognitive apparatus if we are to think about those objects. The possibility of freely choosing between descriptive or non-descriptive thought would undermine the very motivation that made Russell distinguish those ways of thinking in the first place. I call this the Russellian Motivation, that is, the idea that we are epistemically limited beings and that we must exercise our cognitive tools in different ways if we are to grasp certain objects. It was precisely this motivation that led him to distinguish knowledge by acquaintance and knowledge by description7.

Let us take stock. So far I have defined singular thought as non-descriptive thought and showed that this is the most basic definition possible. I also argued that the original motivation for distinguishing singular from descriptive thought was epistemic, i.e., to account for the ways by which our cognition must operate in order to be able to grasp different objects in the universe. I called this the Russellian Motivation. The relevant question now is: should we still keep the Russellian Motivation in mind when discussing singular thought? And the answer is: yes. I see no point in distinguishing singular and descriptive thought if not to account for the same epistemic problem with which Russell was concerned. The distinction is justified if we focus on how we think about things, in how it is possible for our cognition to access the various objects in the universe. This is the interesting problem that must be explained. The motivation can also be put in terms of mental files, singular propositions or any kind of theoretical conception of singular content, as long as we keep in mind that the key question remains about how we can grasp such content. Putting things in terms of content, however, tends to infect the investigation of thought with semantic matters, which is precisely what I am trying to avoid for now. The Russellian Motivation belongs to epistemology, not to semantics. For purely antiseptic reasons, then, I will stick with this more intuitive way of presenting the issue for the moment.

Now, if we accept that the Russellian Motivation is well founded, the question of which objects can be thought about singularly and which require descriptive thinking is immediately raised. Russell already gave his answer: only sense-data and universals (and perhaps the self) can be thought of singularly. Everything else must be thought of descriptively. However, few (if any) philosophers are comfortable with such a restrictive class of singular thoughts nowadays. Most accept that we can have singular thoughts at least about objects in our perceptual field and about objects in our memory. That is to say, most philosophers are willing to weaken Russell"s relation of acquaintance to allow perception of external objects and memory to be counted as acquaintance relations. Such claim is plausible enough, since no description seems to play any role in those cases: there is an intuitive directness about perception and memory that bypass the need of conceptualizations. For this reason, I shall just take perception and memory for granted and focus on the other, more liberal, accounts.

In the next three sections, I will discuss the most common attempts to broaden the class of singular thoughts. I hope to show that none of them is completely convincing, which leaves us with an impasse about singular reference. In the final section, I will argue that this impasse is resolved if we accept a split view of language and thought. For the following discussion, it is useful to follow Russell"s terminology and call acquaintance the sort of epistemic relation that allows singular thought. I will start with the most conservative position within the conservative view, which holds that communication chains are acquaintance preserving.

2.Linguistic Acquaintance

The story goes more or less like this. There is an initial baptism, in which the object being baptized is perceived by the baptizer. The baptizer then passes on the name she introduced in such a ceremony, so that other speakers in her community can use it to refer singularly to the object named. This process goes on and on and eventually ends up reaching me. By learning the name, I then become capable of singularly referring to the object in the same way as the other speakers in the community, and I do so successfully even if I do not know almost anything about the object. I am now one of the links of a communication chain that may have started thousands of years ago. The existence of such communication chain is sufficient to secure singular reference to that precise object.

This is, essentially, Kripke"s causal-historical picture of reference. The conservative, however, claims that not only singular reference is preserved along the linguistic chain, but also the initial acquaintance relation that the original baptizer had with the referent, i.e., perception. We have, thus, acquaintance by testimony, since we rely on the other links of the chain. Put in another way, by being able to think singularly about the baptized object, the baptizer can secure singular thoughts for all other links in the linguistic chain that she initiated, no matter how distant they are. Just by being in such a chain, therefore, I am able to have singular thoughts about the relevant object. Thanks, baptizer!

But how is this possible? Bach (1987) argues that names introduced by perceptual acquaintance function as mental representations of objects, so that when I am thinking through such a name I am representing it non-descriptively in thought. Moreover, I am able to pass along this representation by tokening the name to a hearer, thus she is also able to represent the object singularly in thought by tokening the very name I communicated to her. In any case, what is important about linguistic acquaintance is the idea that names themselves are capable of somehow transmitting the original acquaintance relation to the other links in the chain, so speakers can rely on those links to have singular thoughts in the exact same way that they rely on them to make singular references to that object. The conditions of singular reference, then, are more or less the same as the conditions of singular thought.

Linguistic acquaintance, however, has several problems. First, the argument of communication chains is not an argument about the singularity of reference, but about the preservation of reference and about our general linguistic competence with names, especially with names of remote objects we have very meager information about. As Almog (1984) noted, the role of communication chains is presemantic and has nothing to do with the characteristic semantic contribution that proper names make to sentences in which they occur. What explains the singularity of reference is a general semantic property attributed to the class of expressions called "proper names", a property that can be demonstrated by modal arguments of the sort proposed by Kripke, not by the existence of communication chains. The kind of semantic value that proper names have is not a matter of their ancestry, but of their role in sentences containing them and of the modal profiles of the semantic content they express. In sum, linguistic chains are not in any sense part of the content of proper names, so they do not explain even the singularity of singular reference, let alone singular thought.

Second, names by themselves are not special. Conservatives claim that what distinguishes names introduced via a proper baptism from names introduced merely by description is the existence of a perceptual relation to the referent at the beginning of a communication chain. Perception, then, is the bedrock that guarantees singular thought throughout the community. If perception had no role to play in this case, then any name introduced by description would have the same representational capacity than a name introduced in a proper ceremony just by being a name, and hence it would also secure a singular thought. Of course, this is exactly what the semantic instrumentalism defends and what the conservative view is trying to avoid. Perception, though, is essentially a private and non-transferable thing. It plays an important role in understanding singular terms such as names and demonstratives, yet there is no plausible sense in which it can be communicated by a term. The way the baptizer thinks about the object during the ceremony is hers and hers alone. She cannot, via the transmission of a linguistic device, also display to the hearer her own private epistemic access to the referent. There is no linguistic substitute for perception. In this sense, names introduced in a proper baptism are no better than names introduced merely by description, since what is relevant to infuse them with representational capacities throughout the chain, i.e., perception, is essentially a private thing that simply cannot be passed along to other speakers.

If the foregoing is right, then this type of conservative must either abandon the idea of linguistic acquaintance or find another way to weaken the acquaintance relation. This is what most philosophers do. They argue that public names are indeed not special in any sense; what they do, however, is put us in a causal connection to their bearers, and acquaintance is given by causal connections. I now turn to this idea.

3 Causal Acquaintance

This view apparently avoids the problems raised in the previous section, for their advocates do not resort to any mysterious representational capacity that proper names are supposed to have; causal connections are enough. Also, causal connections can account for the paradigmatic cases of acquaintance as well: perception is a causal connection, so is the memory of a perceptual encounter. More importantly, when I learn a name through a communication chain, I am causally connected with its referent whether it was perceived directly by the baptizer or merely indirectly. In all those cases, in virtue of being causally related to something, I am acquainted with it, and hence I can have singular thoughts about it. Public names are not special in any sense: they are just one of the many possible causal relations that we can bear to objects.

One of the virtues of this view, they claim, is that it accounts for the intuitions of singularity yielded in the following cases that cannot be explained by other theories:

  1. I see some bear prints on the snow and say "wow, this must be a huge one!". It seems plausible to say that I am singularly referring to the bear and having a singular thought about it despite not having perceived it directly. This can be generalized to any case of deferred demonstratives like this.

  2. After some investigation, the detectives from Scotland Yard conclude that the prostitutes murdered in Whitechapel are the work of a single man. The papers coin a name for him: Jack the Ripper. All Londoners then start to refer singularly to the murderer using "Jack the Ripper". It appears that not only the detectives, but also every Londoner can now think singularly about him even without having perceived him or knowing who he is.

  3. Leverrier, after detecting perturbations in the orbit of Uranus, concludes that there must be some celestial body that is responsible for them. He calls it "Neptune" and starts to talk about it using this name. He refers and thinks singularly with it.

In A, B and C what is in fact perceived are only the causal traces of the relevant objects, but it certainly appears that singular reference and singular thought are taking place nevertheless. The reason for this intuition, the conservative claims, is that the object of my thought is determined in virtue of the causal relation itself, and not in virtue of it satisfying certain concepts. There is a fundamental difference between those cases and cases where I merely think of something descriptively, as in "the ugliest man on Earth must be really lonely" or "the shortest spy is a spy": the causal trace of the relevant individual is obviously missing. In other words, just by perceiving the causal evidence left by something is enough to put me in a position to refer and think singularly about it, for it is this very relation that determines which object I am thinking about, not the fact that the object fits a certain description I entertain. This is what Bach called the relational mode of determination of reference8. This theory also explains why we do not have similar singularity intuitions in Kaplan"s Newman 1 case: there is no causal connection for our thought to exploit; after all Newman 1 is not born yet.

Despite its appeal, I remain skeptical about this position. Causal relations do not explain much. We are causally connected with so many things that the motivation for positing acquaintance constraints to singular thought is lost. For example, I am causally related to the designer of McDonald"s logo, whoever he is, just by seeing one of its instances, or to the grandmother of the Vietnamese employee who assembled my cellphone, or to all the stars from which came the carbon atoms in my body. It is even so with objects such as Sagittarius A* - the black hole at the center of Milky Way -, or with Earth"s core. We are clearly in causal connection with all those things, so it follows that we can have singular thoughts about them. In fact, this point can be generalized to all objects that can be detected only through complex instruments and calculations.

Should we really say that we are in a position to think non-descriptively about those objects? The conservative must say yes, for in all those cases I perceive some causal effect that the objects produced, however remote they are and even if their existence is inferred via a very complex procedure. If acquaintance is merely a matter of causal relations, then singular thoughts become cheap. I just need to think about something to which I am causally connected and voilà, I am having a singular thought - nevermind how remote the object is, how many causal intermediaries there are and how complex my thought must be to be able grasp it. This position seems so permissive that we can start wondering if semantic instrumentalism is not a more attractive option.

I am sure that the philosophers who defend such a view do not wish to be so permissive. They may argue that not just any causal connection will do, but only those of the appropriate sort. However, where should we draw the line? When is a causal connection sufficient for warranting singular thought? There are no obvious criteria for stablishing that. If we think of acquaintance as causal connections, then acquaintance becomes a matter of degree: acquaintance relations will be given in a continuum with no evident distinguishing point among its cases. We could classify causal connections as ranging over very strong cases of acquaintance, as in visual perception, to very weak cases of acquaintance, as in the case of the stars which generated the carbon atoms in my body, but not as acquaintance and no-acquaintance, since they are all causal connections. The proposal would still be too liberal. If there is any hope of avoiding this problem, then conservatives must accept that causal connections are just one of the necessary conditions for acquaintance. This, however, weakens considerably their position and its explanatory power, because we can no longer identify the class of acquaintance relations with the class of causal relations. Conservatives must then spell out the other necessary conditions and explain why they are jointly sufficient for acquaintance - a task of tremendous difficulty. In short, they must explain why some causal connections are special so as to count as acquaintance, while others are not. Before that, however, we have no good reasons to simply accept their claim that some causal connection is enough.

Even if a plausible account is given, the conservative view starts to sound as a far cry from the original Russellian picture. The reason is that, even though conservatives are definitely attracted to Russell"s insights, they are nevertheless completely abandoning the Russellian Motivation and stretching the negative definition to the point of rupture. Let me unpack this. Even if I am in an appropriate causal connection with something, it does not follow that I do not need to employ descriptions to think about it. Consider the cases A and B above. When I see the murdered women or the prints on the snow, I must employ a certain description in order to capture the relevant objects, since the only things that are given to my thought non-descriptively are the prints and the victims" bodies. Those are the things that I can cognize without conceptualization. If I want to think about the objects that caused those perceived effects, I must think of them under some description, otherwise my thought would simply not be able to grasp them. I perform an inferential step in both cases, for the objects are not in my immediate surroundings. In order for the conservative view to be coherent, then, the Russellian Motivation must be abandoned, for descriptive thought is clearly playing an essential role in how those objects are captured by the cognizer. This is even clearer with more remote objects, such as Sagittarius A*, which require extremely complex conceptualizations in order to be grasped. Hence, the conservative view is no longer distinguishing singular from descriptive thought in terms of how our cognitive tools must be exercised to capture certain things.

Recanati and Salmon, both defenders of the conservative view, acknowledge this point:

I admit that, by and large, the subject"s access to the referent is mainly descriptive, in the relevant examples. (Recanati 2012, p. 125)

[in the case of a murder scene] The detective need not even be "acquainted" with the murderer, in any ordinary sense; his knowledge of the murderer is by description, in Russell"s phrase. Never mind; he still manages to pull off a de re [singular] belief. It is enough that the believer is appropriately cognitively connected to the person or object. The de re connection need not be direct and intimate; it may be remote and indirect, perhaps consisting of a network of causal intermediaries interposed between the cognizer and the object. (Salmon 2007b, p. 325)

By abandoning the Russellian Motivation, however, the conservative view seems incompatible with the negative definition that was proposed at the beginning of this paper. If descriptions do play a fundamental role in those examples, in what sense can we say that they are cases of non-descriptive thought? This definition, which initially seemed so plausible, clearly no longer captures the whole range of singular thoughts that the conservatives would like to accept. They must either revise it in some way to include those cases of "faux-descriptive" thought or discard it completely.

This is why, I believe, so many different characterizations of singular thought have emerged in contemporary literature. Conservatives felt that there was something right about the distinction between descriptive and singular thought and the general Russellian picture, but they did not feel the pull of the Russellian Motivation. However, without the Russellian Motivation, the distinction between singular and descriptive thought becomes murky. How are we supposed to distinguish those types of thought if not in terms of how our cognition must operate in order to grasp different objects? The conservatives must appeal to other and more sophisticated theories to save the distinction - enter mental files, object-directed thoughts, mental names, etc. The once sharp contrast with descriptive thought, however, becomes blurred.

One common attempt to save the distinction is to argue that, despite the fact that the thinker still must access a causally connected object descriptively, the descriptions employed in those cases are defeasible (e.g. Recanati, 2012). This, they say, is the hallmark of singular thinking. To say that these descriptions are defeasible means that they may fail to apply to the object being thought of: since the object of the thought is already determined by an appropriate causal connection, the descriptive conditions entertained by the thinker may be mistaken without risk of capturing a different object or no object at all. In other words, a singular thought about that precise object takes place even if the object does not fit the description entertained in that occasion. In example A I may successfully think singularly about the creature that left the prints on the snow as "the huge polar bear that was here" even if it was in fact the Abominable Snowman or a person with weirdly shaped feet. This is not the case with purely descriptive thinking, so the distinction is somewhat preserved.

I believe that the conservative"s point about the defeasibility of descriptions is essentially correct: the descriptions that a cognizer uses to think about a causally connected object may be largely mistaken, contrary to cases where the causal relation is absent. However, defeasible descriptions are still descriptions. Why say that the thought with defeasible descriptions is singular rather than a variety of descriptive thought?

Any plausible answer to this question requires the complete rejection of the Russellian Motivation and of the negative definition, because descriptions are still obviously necessary for grasping the relevant object; these thoughts would not be non-descriptive in any straightforward sense. This departure from the Russellian framework may very well be the conservative"s point, and there is nothing obviously wrong with that. However, the price for going down this path is excessively high. If the conservative chooses to characterize singular thought as the thought which employs defeasible descriptions, then she must reject the whole Russellian background that shaped the debate since its beginning, under pain of incoherence. The conservative must also spell out the motivation for distinguishing singular from other types of thought without appealing to Russellian insights about our epistemic limitations. In sum, if the Russellian Motivation and the negative definition are abandoned, then the debate must be cast in different terms with different explanatory purposes than it has been traditionally done. Singular thought, then, would turn out to be something very different than it was at its inception in the literature and with a different role in the philosophy of language and epistemology. But does the phenomenon pointed out above really justify such drastic divorce from Russell? I think not. We could explain it without abandoning the Russellian Motivation or the negative definition. But how should we do that?

Ironically enough, Azzouni (2011) - who is quite liberal concerning acquaintance9 - points the way. Let us take a look at his claims. He says that the defeasibility of descriptions is just a symptom of something deeper:

... the distinction [between singular and descriptive thought] is one between how descriptions are used by the thinker: whether the thinker is focused on the descriptions themselves, or instead focused on the information channels the descriptions are (partially) of. The objects-directed/descriptive distinction, I"m suggesting, is a distinction between a focus on the content of a description through that description as opposed to a focus on (some of) the content of a description independently of that description. (Azzouni 2011, p. 54)

So, for Azzouni, the fundamental feature of singular thought (or objects-directed thought, as he calls it), is the focus on the object being thought of, not in some description or other that it supposedly satisfies. This is why I can afford to misdescribe it: the descriptions are not essential for my thought, and thus are defeasible. In other words, the real distinction is explained by appealing to a Donnellan-style distinction between referential and attributive uses of descriptions. If I employ some description referentially, I am having a singular thought, but if I employ it attributively, I am having a descriptive thought. There is some sense in which I have some object in mind when the description is used referentially (and hence it is defeasible), but it is not so when I think of something as the satisfier of the description.

Now, instead of claiming that this distinction between referential and attributive intentions in the use of descriptions marks the relevant difference between singular and descriptive thought - thus obviously rejecting the Russellian Motivation and the negative definition - we could simply classify this phenomenon as a species of descriptive thought. After all, both attributive and referential uses involve descriptions. In doing so, we can keep the Russellian Motivation, because we account for the fact that our access to the relevant object necessarily involves some description, while also keeping the negative definition unharmed: singular thought remains non-descriptive in the most straightforward sense. It also respects the Russellian Motivation in another manner: the restriction on being able to choose how to think about something. Azzouni"s proposal (and the conservative"s in general) allows the subject some degree of choice in what comes to thinking singularly or descriptively: by learning that some description was largely mistaken, a thinker may choose to have a singular or a descriptive thought depending on the importance she places on the object satisfying the descriptive content. Even if this choice is not entirely free, as in Jeshion"s Cognitivism (a cousin of semantic instrumentalism), it still violates the epistemological limitations Russell was concerned with. In my proposal, some choice is allowed among descriptive thoughts, but not regarding the need of descriptions. What the thinker may perhaps choose is if her descriptive thought was successful or not: if she used some description attributively, then the only object that counts as the intended object is the satisfier of the descriptive content, so her thought would be unsuccessful in case of misdescription; if she used some description referentially, with the intention of thinking about some causally connected object, then misdescription is not a problem. Both thoughts are descriptive nevertheless.

There is one description, however, that cannot fail to apply if our descriptive thought is to be about anything at all: "the object so-related to this causal trace", or something like this. In this case, the object is described in terms of the causal relation itself, and not in terms of other properties it is supposed to have. If this description is not satisfied by anything, then the thinker"s thought cannot land anywhere, precisely because there is no object to land upon. So, even in referential cases, there is at least one description that is used attributively, i.e., the description that is meant to capture the object based on the causal relation itself. Let us call these descriptions relational descriptions10, following Recanati (2010). Descriptions of this kind are what guarantees that the though captures the relevant object even in grossly mistaken referential uses. The same thing applies to causal connections given through linguistic chains. When we learn a name, there is at least one description we must believe to apply to its bearer, as pointed out by Lewis (1979): something like "the one I have heard under the name of N".

If this is plausible, why did so many philosophers feel the need to distance themselves from the Russellian picture if the same phenomenon can be explained by a much more economical theory? My guess is that they believe that this theory hyper-intellectualizes or internalizes descriptions in a very unnatural way11. Nobody seems to think through descriptions of this sort when using public names or making the type of inferences described in examples A, B and C. Phenomenologically, it certainly does not feel as if we are thinking using relational descriptions. These thoughts feel singular. This seems to count as very strong evidence against my proposal. However, our thought is extremely complex. It is very hard to pin down exactly what is going on in those cases. Take indexicals for example. Their characters are given by descriptive rules. Also, indexical sentences have sometimes very complex character building rules. Yet, we do not consciously think about those descriptive rules when using indexicals or indexical sentences. The same applies to something like Perry"s reflexive content. It surely does not feel that we are using and thinking about this content in our ordinary linguistic practices, but it is there nevertheless. If I name the denotation of "the shortest spy" as "Zack", my subsequent tokenings of "Zack" certainly will not feel as involving that description; yet the conservative must consider it a descriptive thought due to the absence of a causal connection. This feeling is not a reliable criterion for identifying singular thought. Glezakos (2009) pointed out that definite descriptions may play a very important role in our ability to wield public names successfully, even though we are not fully aware of the underlying mechanisms by which this happens. Kripke (1980) too acknowledges that speakers often associate a cluster of descriptions with the names they use.

My point here is twofold. First, descriptions are not completely alien to direct reference. We do not need to embrace descriptivism about reference to acknowledge the importance of descriptions for language and thought in general. As long as we keep in mind that singular and descriptive thought are epistemic notions, we will be fine. Second, the objection of internalization presupposes that when we speak or think we are operating with a single content that is completely transparent to ourselves. If Glezakos and Kripke are right to some extent, our thoughts and speech are much messier than this objection assumes. When thinkers are using names or attempting to think about the relevant objects in the examples A, B and C, there is a lot going on that simply cannot be captured by a single description. It is an obvious mistake to think that the description "the celestial body responsible for the perturbations in the orbit of Uranus" was exactly what Leverrier thought when he thought about Neptune. It may have been a different description, or a very complex cluster of descriptions, including a relational one. Maybe Leverrier himself would have a hard time saying what he was thinking.

If this is right, why not say that relational descriptions play a role in all those cases, underlying every referential use of a description? They may not be consciously noticed, but they certainly seem essential for the thinker"s ability to grasp the objects in question. For example, if a thinker is repeatedly corrected about the content of some referential description she consciously used to think about something, she may at one point just say "I"m thinking about the thing so-related to this such-and-such, damn it!" and rest assured that her thought reached the object successfully (if it exists). This is something that everybody could do. The objection of internalization, then, rests on an oversimplified picture of how language and thought work.

I will now briefly discuss the view that discards acquaintance constraints altogether: semantic instrumentalism. If this view is also shown to be unattractive, then it seems that we must be satisfied with allowing singular thoughts only through perception and memory.

4.Semantic Instrumentalism

Semantic instrumentalists hold that the mere manipulation of direct reference devices is enough to guarantee singular thought, so there are no epistemic restrictions whatsoever to singular thinking. When using any proper name or demonstrative, therefore, I am automatically thinking singularly about the referents of those terms. Prominent defenders of this thesis are Kaplan (1989b) and Borg (2004). Kaplan famously said that language itself is capable of generating new thoughts that would be otherwise inaccessible without language; the mere competence with a linguistic tool is enough "to broaden the realm of what can be expressed and to broaden the horizons of thought itself" (Kaplan 1989b, p. 603). Hence, "Acquaintance is an unnecessary artifact, an unwanted relic of a bygone era in the philosophy of language and mind" (Hawthorne & Manley 2012, p 2512). To semantic instrumentalism, then, singular thoughts are really, really cheap. Take any definite description you want; you just have to introduce a name for its denotation and voilà, you are thinking singularly.

Few philosophers are willing to accept such level of liberty, especially when considering cases like Newman 1, where the relevant object not even exists yet. My reasons for rejecting it should be clear by now. Fist, how can a semantic property of a term, i.e., being directly referential, give us non-descriptive cognitive access to an object? Again, any plausible way of answering this question presupposes a rejection of the Russellian Motivation. If thinking via a name is sufficient for having a singular thought, then the kind of epistemic relation that we bear to the object is completely irrelevant. It simply does not matter if my cognition must first reach it descriptively. As I argued earlier, rejecting the Russellian Motivation tends to obscure the very distinction between non-descriptive and descriptive thought, so we should preserve it.

Second, it seems that the distinction of singular and descriptive thought is made irrelevant, because all I have to do to have a singular thought is to introduce a name for the denotation of some description. If this is the case, what is this distinction trying to capture or to explain? It certainly seems otiose. One possible answer is to say that names change the modal profile of the contents of our thoughts: while "Neptune" is rigid, "the celestial body responsible for the pertubations in the orbit of Uranus" is not. Thus, when we introduce names, we "broaden the horizons" of thought in the sense that we become able to entertain referential contents with distinct modal behaviors. This seems to be the main pull of semantic instrumentalism13.

It is true that the usefulness of introducing descriptive names like "Neptune", "Jack, the Ripper" or "Vulcan" is to allow counterfactual reasoning without capturing the wrong truth conditions. Also, descriptive names seem to function as cognitive shortcuts; we token them in thought without explicitly thinking about the descriptive conditions we employed to fix their reference. However, this does not mean that our cognition is able to "reach" the object non-descriptively just because we do not consciously think about some description or other while tokening them. We cannot explain or give the meaning of a descriptive name by pointing at something; we must describe its bearer. Just because we do not "feel" to be thinking descriptively does not mean that we are not. To say so would imply the rejection Russellian Motivation, which we are trying to keep.

5.The Split View

As we saw, we have good reasons to doubt these attempts to extend singular thoughts beyond the paradigmatic cases. But where does this leaves us regarding singular reference? I argued that, if we accept the Russellian Motivation, then we should say that we think descriptively in many cases in which we use public names and demonstratives, which are devices of singular reference. This leads to an impasse. Should we say that the references made with these devices are also descriptive, contrary to what is normally believed? If we say yes, then we must accept some sort of descriptivism about names and demonstratives, which seems wrong14. But if we say no, then we must admit a split between language and thought: while reference remains singular, our thoughts about the referent are descriptive. For ease of exposition, I will now speak in terms of singular propositions, and assume that having a singular thought is grasping a singular proposition.

The split between language and thought is repulsive to many philosophers. The reason behind this reaction is that those philosophers hold what I call the mirroring intuition, i.e., the intuition that the semantic content of language must somehow mirror or reflect the content of thought. After all, if we can assert a singular proposition then it seems obvious that we can also have other attitudes towards that same proposition, such as belief and knowledge, which require grasping it. In other words, if we are able to refer singularly, then it must be the case that we are also able to think singularly. If we accept this split, however, it would be possible for a speaker to express a proposition that she herself does not understand15. To those who have the mirroring intuition, this idea is absurd16. It certainly was to people like Frege and Russell.

My proposal here is even more dramatic: I claim that this is not only possible, but that we do it systematically. We express singular propositions that we are unable to grasp very often. This is exactly what happens in the examples A, B and C, with all proper names whose bearers we did not perceive and also with deferred demonstratives. It even happens with the apparently innocent indexical "tomorrow": every utterance of it express a singular proposition that we cannot grasp simply because its referent does not exist at the time of the utterance, exactly like the Newman 1 case; it can only be thought of descriptively.

Why should we be bothered by this? The singularity of reference is a thesis about the characteristic semantic contribution that some referring expressions make to the truth conditions of the sentences in which they appear. This characteristic contribution is a semantic property attributed to classes of expressions by the rules of language, to the effect that all expressions of that class, when uttered, will have such and such semantic value with such and such modal profile. Singular terms express singular contents. This is a fact of language. Expressing a singular proposition, then, is merely a matter of uttering a sentence containing a singular term in accordance with the linguistic norms. In other words, referring singularly is an act that can succeed or fail given facts about the external world and linguistic conventions. The way we think about the referent has no bearing in that matter. Thus, there is nothing inherently absurd about expressing a singular proposition without being able to grasp it.

This becomes even clearer when we appreciate the fact that singular propositions are not purely linguistic or conceptual entities. They are partly constituted by extramental objects. As such, there is no reason to assume that mere linguistic knowledge is enough to grasp them. Putting language aside, we normally say that it is possible to cognize an object non-descriptively only if it is perceptually accessible. If it is implausible to claim that we can cognize an object non-descriptively without a perceptual encounter, why claim that we do cognize objects non-descriptively as soon as they appear in discourse, dressed in words? Why should the conditions for cognizing a singular proposition be different than the conditions for cognizing non-descriptively the extramental object that is its ingredient? A singular term is not an epistemic shortcut to its referent. It is a linguistic device for expressing singular contents, i.e., a tool for making singular references. If singular terms have worldly objects as semantic values, grasping these values is not something that can be achieved simply by understanding linguistic rules: they are out there, in the world, and some substantial epistemic relation is needed to capture them. Expressing singular propositions, then, is a matter of appropriately exploiting the linguistic conventions; what is required for grasping or understanding them is a whole other matter.

It could be argued that grasping the propositions we express is a necessary condition of our linguistic competence. If we did not grasp them, how could we explain our proficiency with singular terms referring to remote objects? There is a simple answer to that. When we learn language, we learn how to use singular terms in certain paradigmatic cases, where the referent is perceptually present. We therefore learn what singular terms are used for, how to interpret them in sentences we hear and what kind of semantic content they typically express. It is not hard to extend this ability to non-paradigmatic cases. In fact, for many communicative purposes, referring to an object and knowing that we were successful in doing so is enough. When we talk about Aristotle, for example, we know that we refer to him by using the public name "Aristotle", and we know that we express a singular proposition containing Aristotle as its constituent. For our communicative purposes, however, we do not need also to grasp the singular proposition being expressed. It is enough to know what role the name "Aristotle" is playing in our utterances and to think descriptively about its bearer. Similarly, when I say "this must be a huge one!" pointing to bear prints on the snow, I may know that my singular reference to the bear was successful (and so may the audience), but I do not need also to grasp the singular proposition I expressed (and neither does the audience) for the communication to succeed. If we could grasp it, so much the better; in many cases, however, knowing what role the singular term is playing and under which conditions the utterance is true is as good as it gets.

We must also be careful in what we mean by content. If we mean only the referential truth conditions of thoughts and utterances, then they will indeed have the same content, and the mirroring intuition would be in force. To explain the way we think about things, however, is not enough to invoke this kind of content, for the referential content is determined by linguistic rules and facts of the world, not by our cognitive operations. Surely we can attribute singular contents to cognizers as a way of describing and classifying their actual thoughts in terms of their referential truth conditions. This does not mean, however, that these thoughts are in fact singular, i.e., that the thinker thinks about the referent non-descriptively. Reports of this kind are often done from a privileged point of view17: we, as reporters, know what objects are relevant for the truth of the thoughts in question, so we can ascribe a singular proposition to a thinker even when she is not in fact having a singular thought. This is useful for many purposes, including theorizing about language and thought. What matters in these cases is just that we get the objects the thought is about correctly. In short, singular propositions may be attributed even when the actual thought in the thinker"s head is descriptive. Thus, we must distinguish referential content from content in general; the first is what we often have in mind when reporting or describing someone"s actual thought in terms of its referential truth conditions, and the latter is the actual thought. The explanation for how thinkers think about the objects they refer to, therefore, is not found at the level of the referential content, but at the level of the actual thought. What makes a thought singular or descriptive is the way the thinker thinks about the referential content, not this content itself.

Russell obviously held the mirroring intuition. Because of this, and because he had the Russellian Motivation in mind, he had no choice but to consider ordinary public names (and deferred demonstratives as well, I suspect) as disguised descriptions: since our cognitive access to their referents must be descriptive and given that the semantic content of expressions should reflect the contents of our thought, then it follows that those expressions must semantically function as descriptions, not logically proper names. Therefore, they do not express singular propositions. When facing the same problem, what most philosophers do is drop the Russellian Motivation while keeping the mirroring intuition. In this way it is possible to maintain that we refer singularly with any singular term, but acquaintance must be extended or abandoned completely, or else many propositions we express would be inaccessible to us. Rejecting the Russellian Motivation, however, leads to many difficulties in distinguishing descriptive from non-descriptive thought effectively.

Instead, we should keep the Russellian Motivation and drop the mirroring intuition. The distinction between singular and descriptive thought then keeps sharp, we respect the main reasons why it was introduced in the first place, and still explain the same phenomena. What is a relic of a bygone era in the philosophy of language is not acquaintance, but the mirroring intuition. When abandoned, we can appreciate that we refer singularly without entertaining singular thoughts more often than we believed. Singular reference is cheap, while singular thought is not.


ALMOG, J. "Semantical Anthropology", Midwest Studies in Philosophy, IX, 479-489, 1984. [ Links ]

______ "Is a Unified Description of Language-and-Thought Possible?", The Journal of Philosophy, Vol. 102, No. 10, 493-531, 2005. [ Links ]

______ "Referential Uses and the Foundations of Direct Reference". In Almog and Leonardi (eds). Having in Mind: The Philosophy of Keith Donnellan (pp. 176-184). Oxford: Oxford University Press, 2012. [ Links ]

AZZOUNI, J. "Singular Thoughts (Objects-Directed Thoughts), Proceedings of the Aristotelian Society Supplementary Volume lxxxv, 45-61. 2011. [ Links ]

BACH, K. Thought and Reference. Oxford: Oxford University Press, 1987. [ Links ]

______ "Putting a Thing into a Thought". In New Essays on Singular Thought (pp. 39-63). Oxford: Oxford University Press, 2010. [ Links ]

BORG, E. Minimal Semantics. Oxford: Clarendon Press, 2004. [ Links ]

______ "Pointing at Jack, Talking About Jill", Mind & Language, Vol. 17, Issue 5, 489-512, 2002. [ Links ]

GLEZAKOS, S. "Public proper names and idiolectal identifying descriptions", Linguistics and Philosophy. 32:317-326, 2009. [ Links ]

HAWTHORNE, J. & MANLEY, D. The Reference Book. Oxford University Press, USA, 2012. [ Links ]

JESHION, R. "Donnellan on Neptune", Philosophy and Phenomenological Research Vol. LXIII, No. 1, 111-135, 2001. [ Links ]

______ "Singular Thought: Acquaintance, Semantic Instrumentalism, and Cognitivism" in New Essays on Singular Thought (pp. 105-140). Oxford: Oxford University Press, 2010. [ Links ]

KAPLAN, D. "Demonstratives", in Almog, Perry and Wettstein. (eds.). Themes from Kaplan (pp. 481-564). Oxford: Oxford University Press, 1989a. [ Links ]

______ "Afterthoughts", in Almog, Perry and Wettstein. (eds.). Themes from Kaplan (pp. 565-614). Oxford: Oxford University Press, 1989b. [ Links ]

KRIPKE, S. Naming and Necessity. Harvard University Press, 1980. [ Links ]

LEWIS, D. "Attitudes De Dicto and De Se". The Philosophical Review, Vol. 88, No. 4, 513-5431979, 1979. [ Links ]

RECANATI, F. "Singular Thought: In Defense of Acquaintance". In New Essays on Singular Thought. (pp. 141-190). Oxford: Oxford University Press, 2010. [ Links ]

______ Mental Files. Oxford: Oxford University Press, 2012. [ Links ]

RUSSELL, B. "Knowledge by Acquaintance and Knowledge by Description". In Mysticism and Logic (pp. 209-232), London: Unwin, 1949. [ Links ]

SALMON, N. "Demonstrating and Necessity". In Content, Cognition, and Communication (pp. 67-99). Oxford: Clarendon Press, 2007a. [ Links ]

______ "The Good, the Bad and the Ugly". In Content, Cognition, and Communication (pp. 309-339). Oxford: Clarendon Press, 2007b. [ Links ]

SCHIFFER, S. "Descriptions, Indexicals, and Belief Reports: Some Dilemmas (But Not the Ones You Expect)". Mind, New Series, Vol. 104, No. 413, 107-131, 1995. [ Links ]

18CDD: 153.42

* Many thanks to Nathan Salmon, Marco Ruffino, Kepa Korta, Lucas Miotto and to two anonymous referees for the many valuable comments and suggestions. I also thank Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), which funded my period as a visiting scholar at UCSB, where this paper was written.

1Note, however, that the philosophers I call 'conservatives' might be quite liberal when compared to the more stringent Russellian view of acquaintance. They are conservatives because they oppose the so-called semantic instrumentalism.

2Which I agree.

3An example of this strategy is Bach (1987).

4An anonymous referee pointed out that one cannot tell by the mere occurrence of a description in a sentence that the thought it expresses is really descriptive, for the description might be being used referentially and not attributively. In a certain sense this is true. In typical referential uses, the object being referred to is already perceptually given (as in all Donnellan's famous examples). Assuming that perception allows for singular thinking (as I do in this paper; more on this below), a singular thought is already guaranteed by perceptually attending to the object. In other terms, the description I use is superfluous to my thought, for the object is already determined to my cognition; the description I utter is just an aid for communication, an useful tool to direct audience's attention to the object I already have in mind (Cf. Almog, 2005; 2012). So, in these cases, the occurrence of a description in a sentence is not an infallible indicator of a real descriptive thought, as the referee correctly claims. However, we must be careful not to confuse thought with reference here. Even though my thought is singular, it does not immediately follow that I also express a singular proposition with the description I utter. This would be already to assume the so-called semantic significance of Donnellan's distinction, i.e., the thesis that descriptions used referentially express singular propositions (or directly refer) at the level of their official or literal content. If this thesis is false (as I think it is), a description will always express a general content regardless of it being used referentially or attributively; it does not matter if my thought is singular or descriptive when I use it. In this case, Donnellan's distinction would belong to pragmatics, not to semantics. Of course, the thesis of semantic significance may be true, but it must be argued for and not simply concluded from the fact that we use descriptions referentially and attributively. The upshot of this reasoning (and the general point of the paper, in fact) is this: to know if a thought is singular or not, look to the kind of epistemic contact the cognizer has with the object, not to the expressions she uses to refer to it.

5See Jeshion's introduction in Jeshion (2010), Recanati (2010) and (2012), Azzouni (2011) and of course Bach (1987), to name a few. They all oppose singular thought to descriptive thought in one way or another.

6Emphasis mine. Although he talks about proper names, we can easily generalize the point to thought. For Russell, singular reference and singular thought have exactly the same epistemic requirements, i.e., acquaintance.

7Of course, Russell's epistemological theses are closely connected to his views on semantics and linguistic understanding. Nevertheless, I believe that epistemology has some primacy here: it is precisely because we cannot have acquaintance with remote objects that their names cannot be genuine names, for example; they must function like descriptions, otherwise we would not be competent with them. As it seems clear, for Russell, epistemology determines semantics, so to speak. In other terms, how we think about things dictates the semantics of the expressions we use to express our thoughts about them. In fact, I believe that this primacy of epistemology over semantics was the dominant view in philosophy of language until the rise of direct reference and content externalism, and even now it still has great appeal. See, for example, how Frege's Puzzle and the phenomena of cognitive value still haunt semanticists, even those sympathetic to direct reference and externalism (e.g. Kaplan and Perry). Thanks to an anonymous referee for making me flesh this out more precisely.

8 Bach (1987) p. 12, and also Recanati (2012) p. 21 and p. 34. Note that Bach is skeptical about generalizing this sort of relation beyond communication chains, although he does not assertively deny it. He is an advocate of the 'linguistic acquaintance', as discussed in the previous section. He is very clear on this point in Bach (2010).

9 Azzouni (2011) is not a conservative and (apparently) not a semantic instrumentalist either. His view seems to lie somewhere in the middle between those positions, since he rejects stringent restrictions on singular thinking but apparently does not embrace semantic instrumentalism.

10Lewis and Russell himself advocate something like relational descriptions. However, they accept a full-blown descriptivism about thought (and reference), in which relational descriptions of the kind mentioned here are employed even in perceptual cases. In other words, in their view, descriptions are necessary to capture even objects that are in the thinker's perceptual field. I think that this is mistaken. We can accept relational descriptions only for cases where it seems obvious that we need descriptions to be able to grasp an object.

11 Recanati (2010) and (2012) express precisely this concern.

12Hawthorne and Manley hold what they call 'liberalism', which is very similar to semantic instrumentalism. The relevant contrast is that they believe that the existence of acquaintanceless singular thoughts is just a brute fact and provide no positive arguments about how they are possible. Thanks to an anonymous referee for pointing this out.

13See Recanati (2012) chap. 13.

14 Schiffer (1995) argued along these lines, saying that demonstratives may be used attributively, thus expressing a general proposition. Borg (2002) and Salmon (2007a), on the other hand, offer very compelling arguments about how deferred demonstratives function in the exact same way as demonstratives used perceptually. The semantic rules of demonstratives are general enough to cover both regular and deferred cases.

15Jeshion expresses precisely this concern in print in Jeshion (2001). I have also heard this objection many times in conversation.

16Salmon is a noteworthy exception. He argues that being unable to grasp a proposition is no impediment to asserting it. As he says, assertion demands very little of the speaker's cognition. So, despite being a conservative about singular thought, he does not accept the mirroring intuition, at least not in the same way as others do. See Salmon (2007b).

17See Almog (2005) for a detailed discussion on this topic.

Received: October 15, 2015; Revised: January 05, 2016; Accepted: February 23, 2016

Creative Commons License This is an open-access article distributed under the terms of the Creative Commons Attribution License