Frege's views on propositional constituency are intimately related to his basic semantic picture.  Fregean senses are, at once, the ingredients in propositions, and the items in terms of which the word-world relation is to be understood.  Russell's idea that the references of expressions can be constituents of propositions, an idea that sounds bizarre to those of us brought up on Frege, is very much connected to his rejection of Frege's sense-reference view and his substitution of a very different basic semantic perspective.⁽²⁾

There is another aspect of Russell's view about propositional constituents worth noting here.  Focusing upon this aspect will set the stage for a discussion of just how different Russell's basic semantic picture is from Frege's. It will illustrate, moreover, the intimate connection between Russell's semantic views and his epistemological outlook. Russell announces a "fundamental epistemological principle":

Every proposition which we can understand must be composed wholly of constituents with which we are acquainted. ("Knowledge By Acquaintance and Knowledge By Description," p.211)

Russell goes on to explain that

we have acquaintance with sense-data, with many universals, and possibly with ourselves [Russell here has Humean worries about the self as a possible object of acquaintance], but not with physical objects, or other minds. ("Knowledge By Acquaintance and Knowledge By Description," p. 223)

Russell's fundamental principle doesn't quite limit the constituents of propositions to objects of direct acquaintance. There may be propositions that contain things with which we are not acquainted.  These, however, are not propositions that we can understand or, presumably, assert.  The propositions that we can understand or assert, then, may contain only sense data, universals, and (perhaps) ourselves.

The dispute between Frege and Russell concerning the constituents of propositions, I have said, betokens a fundamental difference in semantic perspective. Russell maintains that when one is acquainted with something, say, a present sense datum or oneself, one can refer to it without the mediation of anything like a Fregean sense. One can refer to it, as we might say, directly.⁽³⁾  Indeed, Russell invents a new linguistic category, the "genuine" or "logically proper" name, that subsumes expressions that function in this most unFregean way.  A logically proper name--Russell's examples of logically proper names of particulars are `I' (if indeed we are acquainted with the self) and `this' (when the latter is used to refer to a present sense datum)--is a

mere noise or shape conventionally used to designate a certain [thing]; it gives us no information about that [thing], and has nothing that can be called its meaning as opposed to denotation. ("Knowledge By Acquaintance and Knowledge By Description," p.218)

I have so far called attention to two Russellian notions that depart radically from the Fregean outlook: nonconceptual constituents of propositions, and conceptually unmediated reference.  Acquaintance, as we have seen, furnishes a link between them.  One can refer directly, by means of a genuine name, only to an object of acquaintance, and when one does so, the referent itself will be a constituent of the proposition expressed.

Russell's idea of direct, Sinnless reference, is one that Frege does not ever explicitly consider, and one that he surely would have rejected, both early and late.  Frege, in the Begriffsschrift, seems, however, to have advanced a closely related idea. Frege appears to have held that while names have associated modes of presentation--he calls them "ways of determining the content"--the semantic function of a name is not to introduce its associated mode of presentation into the proposition expressed. The semantic function of a name is rather merely to make its referent a subject of discourse.⁽⁴⁾  The early Frege thus shares with Russell the idea that nothing sense-like gets into the proposition.  Russell's proposal, however, goes far beyond Frege's early view.  Russell, unlike Frege, gives no role at all to modes of presentation. The connection between linguistic expression and referent, for Russell, is not to be explained by the referent's fitting the term's associated mode of presentation, the referent's having the property associated with the term, for there is no such mode of presentation, or property, associated with the expression.

What was so unthinkable to Frege about Sinnless reference? It is sometimes suggested that Frege rejects the possibility of direct reference because of his puzzle about informative identity. This seems unlikely. The Frege of the Begriffsschrift, I've suggested, had modes of presentation in his intellectual repertoire. The early Frege didn't suppose, though, that this solved the problem of informativeness.  Still, one might think, senses were at least necessary (if not sufficient) for a solution. I don't believe that even this much is correct. Frege's Begriffsschrift metalinguistic account of identity sentences works just as well for sense-less, directly referential names as it does for names that have associated modes of presentation.  Let `Hesperus' and `Phosphorus' be directly referential, mere "noises or shapes conventionally used to designate," that give no information about their referents.⁽⁵⁾  Given Frege's metalinguistic account of identity, the proposition expressed by 'Hesperus = Phosphorus' will be the nontrivial proposition that the referent of the one name is identical to the referent of the other, quite different name.  Thus Frege's puzzle is not, at least to the early Frege's mind, solved by the introduction of senses, and his proposed solution would work as well for Sinnless names.  All of this suggests that the informativeness puzzle was not the key to Frege's adoption of his fundamental thesis that reference requires an associated mode of presentation.

One gets the feeling from reading and re-reading Frege, moreover, that his implicit denial of the possibility of conceptually unmediated reference plays a role much more fundamental than is allowed for by the prior suggestion. It is not merely that the reference-without-sense picture is not adequate to certain puzzles. It is that the very idea of Sinnless reference is somehow incoherent.  In some strong sense, there could not be reference without sense.  Whether or not Frege held this strong view, and I bet that he did, it is certainly consonant with the things that he does say.  It is, moreover, a view worth exploring, if for no other reason because it is the sort of thing one often hears in discussions with philosophers of broadly Fregean orientation.  Why might Frege, or anyone else for that matter, think that Russell's idea is thus incoherent, or impossible, or something of the like? 

My speculation is that something very deep inclined Frege to suppose that reference without sense was impossible, Frege's philosophical picture of thought and of the contents of thought. Here's an analogy. A view sometimes called "the representative theory of perception" maintains that the perception of a physical object requires, perhaps consists in, the apprehension of sense-data that represent the physical object. Direct perceptual apprehension of physical objects, on such a view, is impossible, perhaps even incoherent. My proposal is that Frege held an analogous view about thought, a "representative theory of conception." The idea here is that thought about an object requires, and perhaps consists in, the apprehension of a concept that represents the object. Frege, I want to suggest, was drawn, even early along, to the idea that thinking of an object requires--maybe even consists in--the direct apprehension of a sense, of a mode of presentation of the object.

This traditional idea has had considerable appeal. Indeed, what is taken to be the alternative might seem almost magical: the direct apprehension in thought of things thought about. When you think about me, to give an example, don't you have to think about me in a certain way, for example, as the author of this paper?  What would it mean to just think about me, to think about me without, so to speak, bringing me under a concept?⁽⁶⁾  Such considerations, whether or not they are in the end conclusive, have certainly made the representative theory tempting, prima facie plausible.

If Frege, even at the time of the Begriffsschrift, indeed accepted the representative theory of conception, this would explain why reference-without-sense would seem incoherent. Notice that to accept the representative theory, one need not accept Frege's later view that the representations constitute the content of the proposition expressed. The propositional content might even include--as Russell maintained, and as Frege's remarks in the Begriffsschrift suggest--the object thought about. There is, of course, a certain tension between this Russellian conception of propositional content and the idea that to be thinking of Hesperus is just to be entertaining a concept. It is this tension, I want to suggest, that gets resolved with Frege's "On Sense and Reference" conception of propositional content.

When one stands in direct epistemic contact with something, according to Russell, one needs no representational intermediaries in order to think about, or refer to, it. Russell was, in this respect, an arch-anti-representationalist.  Notice, however, the role of the epistemic immediacy.  It is what precludes the need for any representational intermediary.  What about examples in which someone speaks or thinks about something with which he is not acquainted?  Here Russell's representationalism comes to the fore. I cannot, held Russell, directly apprehend external objects, the pen I am now using, for example. The only way I can make cognitive contact with such things, albeit a kind of inferior contact, is (roughly) by apprehending concepts that represent the things.  I say "roughly" because Russell's conception of these things is so different from Frege's that talk of Russellian "mediating concepts" needs substantial qualification--to be provided in Section V below. Russell, in any case, although--as I will temporarily put it--he endorsed a kind of conceptually mediated reference, he never made it ubiquitous, as did Frege. We revert to mediating concepts only when we wish to speak of something with which we lack epistemic intimacy.

This is not to say, however, that Russell took conceptually mediated reference to be a kind of unusual exception.  The overwhelming majority of the things about which people ordinarily speak, Russell is the first to admit, are not objects of acquaintance.  Ordinary names of other people, since they refer to things with which I cannot be acquainted, surely don't function for me as genuine names, but rather as concept introducers. Ordinary proper names, that is, function as definite descriptions. Even demonstratives like 'this' and 'that', expressions that intuitively perhaps come as close as any to being pointing devices that do not refer by conceptually characterizing the referent, are ordinarily used not to refer to sense data, but to external physical objects.  Most ordinary singular terms, according to Russell, are surrogates for definite descriptions. Logically proper names, despite their theoretical interest, don't seem to have much to do with ordinary linguistic practice. 

Common words, even proper names, are usually really descriptions.  That is to say, the thought in the mind of a person using a proper name correctly can generally be expressed explicitly if we replace the proper name by a description.  Moreover, the description required to express the thought will vary for different people, or for the same person at different times.  The only thing constant (so long as the name is rightly used) is the object to which the name applies.  But so long as this remains constant, the particular description involved usually makes no difference to the truth or falsehood of the proposition in which the name appears. (P.208, "Knowledge By Acquaintance and Knowledge By Description")

Sounds just like Frege, no?

In the case of an actual proper name such as `Aristotle' opinions as to the sense may differ.  It might, for instance, be taken to be the following:  the pupil of Plato and teacher of Alexander the Great.  Anybody who does this will attach another sense to the sentence `Aristotle was born in Stagira' than will a man who takes as the sense of the name: the teacher of Alexander the Great who was born in Stagira.  So long as the reference remains the same, such variations of sense may be tolerated, although they are to be avoided in the theoretical structure of a demonstrative science and ought not to occur in a perfect language. ("On Sense and Reference," p.58, footnote)

We seem to have arrived at a Frege-Russell consensus.  Ordinary singular terms, at least the great majority of them, are not directly referential.  They refer by expressing concepts. This is not, of course, to say that Frege and Russell agree on all the essentials.  Russell, but not Frege, thinks that at least some singular terms are directly referential.  Russell's reason, moreover, for thinking that the great majority of singular terms express concepts is that the great majority of the things about which we wish to speak are not objects of acquaintance.  Frege certainly never suggests that epistemic immediacy with an object would somehow make possible Sinnless reference to it. Finally, as noted above, talk of Russellian "mediating concepts" needs serious qualification.

Nevertheless, Frege and Russell seem to share something of a common perspective. Ordinary singular terms make mediated semantic contact with things. The propositional constituents that corresponds to ordinary singular terms, moreover, is, for both Frege and Russell, the mediating representation, and not the reference.  Russell, one might suppose, was the first in a long line of neo-Fregeans.

Russell, however, seems to take pains to distance himself from Fregean representationalism. His view, contrary to the suggestion of the last paragraph, was not that Frege's approach, or something much like it, successfully accounts, at least in general, for the linguistic phenomena; that such an approach misfires only with regard to the theoretically interesting, although rarely occurring, phenomenon of logically proper names.  Russell believed, on the contrary, that the sense-reference picture was fundamentally mistaken.⁽⁷⁾ How then shall we construe his representationalism?  How, specifically, does it differ from Frege's?

Let us begin by drawing a distinction, crucial from Russell's point of view, between two sorts of definite descriptions, ones like `the most long-lived of men', as opposed to ones like `the material object causally responsible for this' (where `this' refers to a present sense-datum).⁽⁸⁾  The relevant difference is that only the first is, as we might say, "purely qualitative," or "purely descriptional."⁽⁹⁾ The second, insofar as it contains what Russell would consider a genuine name, is a sort of hybrid.  Russell held that the definite descriptions our ordinary singular terms conceal are not purely descriptional.

. . . a description known to be applicable to a particular must involve some reference to a particular with which we are acquainted, if our knowledge about the thing described is not to be merely what follows logically from the description.  All names of places--London, England, Europe, the earth, the Solar System--. . . involve, when used, descriptions which start from some one or more particulars with which we are acquainted.⁽¹⁰⁾  I suspect that even the Universe, as considered by metaphysics, involves such a connection with particulars.  In logic, on the contrary, where we are concerned not merely with what does exist, but with whatever might or could exist or be, no reference to actual particulars is involved. ("Knowledge By Acquaintance and Knowledge By Description," p.210)

When I say, "London is ugly," or, to use an indexical phrase instead of a name, "This pen ran out of ink," the respective singular terms, `London' and `this pen', purport to refer to things that are not possible objects of acquaintance, and so, thinks Russell, those terms must be surrogates for definite descriptions.  `London' and `this pen' are, at the same time, terms that I "know to be applicable to particulars," and so the descriptions that each of these terms conceal must involve reference to things with which I am acquainted.  The description concealed by `this pen' might be, for example, `the pen causally responsible for this' (where the demonstrative refers to some relevant sense datum). 

Russell's notion of genuine name thus plays a role much more central than we had supposed.  It is not merely that Russell makes theoretical room for direct reference, even as he admits that, in practice, ordinary singular terms are, by and large, descriptional. Direct reference is involved even in our conceptually mediated references to ordinary things.  Russell's representationalism is thus a far cry from Frege's.

What led Russell to make the hybrid description the effective paradigm, to reject the Fregean purely qualitative paradigm? As noted above, Russell tells us that

. . . a description known to be applicable to a particular must involve some reference to a particular with which we are acquainted, if our knowledge about the thing described is not to be merely what follows logically from the description.

There is, Russell seems to be saying, a certain sort of knowledge about the denotation of a purely qualitative description that is, in principle, unattainable. So, if ordinary names abbreviated such descriptions, we could not have this sort of knowledge of their denotations.  Russell holds, however, that, in the case of ordinary names, we surely do have this kind of knowledge. 

One thing that is confusing about Russell's discussion is that he has characterized the impossible-to-attain knowledge in two ways that don't sound like they come to the same thing. We cannot know, he tells us, that a purely qualitative description is "applicable to a particular." And we cannot know "anything that does not follow from the description itself."

Let's begin with the second, and more tractable, formulation. One can know, perhaps one automatically knows, that the denotation of `the most long-lived of men', if such exists, must be a man, and must have lived longer than any other man.  This much presumably "follows from the description." One cannot go beyond such knowledge, however, and come to know that the denotation of this description is, say, Chinese, or is the man standing before me now. (Why not? I'll return to this below.)

What about Russell's other characterization of the unattainable sort of knowledge, "knowledge that the description is applicable to a particular"?  Russell apparently does not here intend to rule out the possibility of coming to know that a purely qualitative description has a denotation, that something or other satisfies the description.  He tells us, for example, that

. . .  `the most long-lived of men' is a description which must apply to some man, but we can make no judgements about this man which involve knowledge about him beyond what the description gives. ("Knowledge By Acquaintance and Knowledge By Description," p. 210)⁽¹¹⁾

The knowledge Russell apparently means to exclude as "knowledge that the description is applicable to a particular" is knowledge that, as Russell sometimes says, the description applies to a "particular particular." One cannot come to know, that is, to which thing it is that the description applies. One cannot, for example, come to know who was the most long-lived man.⁽¹²⁾ 

The case is quite different, according to Russell, for hybrid descriptions, or names that abbreviate the latter.  I can know not only that the name `Jonathan Wettstein', the name of my son, has a denotation. I can also know which thing it is to which the name applies.  It applies, as noted, to my son, the very person sitting across the table from me now, and so on.⁽¹³⁾

Putting aside the differences in the two formulations, Russell certainly seems to be saying that we cannot, in principle, come to know anything substantive about the denotation of descriptions like `the most long-lived of men'. It's time that we asked: "Why in the world not?" Why, in principle, couldn't we conduct a study, and come to learn that there is a uniquely longest-lived person, that she is Chinese, that her name is such-and-such, and so on.  Indeed, allowing ourselves Russell's intuitive talk of knowing who, we could presumably meet the relevant person and thus come to know who she is, for example that she is the person standing before me now.  Why should, and how could, the fact that our initial specification was in purely qualitative terms preclude such knowledge?

I suspect that Russell has not expressed himself well, that he would have agreed that we can have knowledge of the denotation of a purely qualitative description that goes beyond the information contained in the description.  Russell's motivation for the thesis that ordinary names are descriptional, but not purely descriptional, is not, I want to suggest, quite what it appears to be.

Russell tells us in The Problems Of Philosophy that

. . . there are various stages in the removal from acquaintance with particulars: there is Bismarck to people who knew him; Bismarck to those who only know of him through history; the man with the iron mask; the longest-lived of men.  These are progressively removed from acquaintance with particulars; the first comes as near to acquaintance as is possible in regard to another person; in the second, we shall still be said to know `who Bismarck was'; in the third, we do not know who was the man with the iron mask, though we know many propositions about him which are not logically deducible from the fact that he wore an iron mask;⁽¹⁴⁾ in the fourth, finally, we know nothing beyond what is logically deducible from the definition of the man. (P. 58)

Let's distinguish, along Russellian lines, three ways of making epistemic contact with particulars.  At one extreme, one makes the most direct sort of epistemic contact. Bismarck, for example, is acquainted with himself.  At the other extreme, one makes direct contact not with a particular, but with a constellation of purely qualitative universals, the sort of thing expressed by a purely qualitative definite description.  If such a constellation of universals is uniquely instantiated, say by an external material object or another person, then one might be said to be making a kind of inferior, mediated epistemic contact with that thing. 

There is an intermediate case. One might make direct contact with a constellation of universals and acquainted-with- particulars, the sort of constellation expressed by a hybrid description like `the person who is my only daughter' (say, as used by me). Such a constellation affords me a mediated but still intermediate level epistemic access to my daughter, for she is being presented as related to a particular with which I am directly acquainted, in this case, me.

It's crucial to note here that Russell links this intermediate epistemic access to knowing who.⁽¹⁵⁾ Not only Bismarck's friends, but even someone who knows Bismarck only through history, can be said to know "who Bismarck was."  Russell's idea here, I think, is that if I am competent with the name `Bismarck', then either I've met Bismarck and have been introduced to him by name, or I've learned to use the name by interacting with others who use it.  Either way, I stand in the intermediate epistemic position discussed above. That is, I am acquainted with a constellation of universals-cum-particulars that together specify Bismarck.  If I've met Bismarck, the relevant particulars with which I'm acquainted are "certain sense-data which [I] connect (rightly, we will suppose) with Bismarck's body." (P.209) For those of us who haven't met Bismarck, the relevant objects of acquaintance consist in "testimony heard or read."⁽¹⁶⁾ In either case, thinks Russell, as in the case of "my daughter," I have an epistemic fix strong enough for it to be reasonable to say that I know who is in question.⁽¹⁷⁾

Russell maintained that ordinary names were surrogates for hybrid descriptions, descriptions that provide intermediate level epistemic contact with the denotations. One who is competent with an ordinary name, he further held, knows who it names. It begins to look as though Russell took there to be a strong epistemic requirement for the use not only of genuine names, but even of ordinary names. The use of ordinary names, no less than genuine ones, requires identifying knowledge on the part of the competent speaker. One who uses a name, genuine or ordinary, must know about whom he is speaking. One who uses a genuine name must of course satisfy a much more stringent epistemic requirement, than need one using an ordinary name. Only the former requires direct acquaintance with the referent.

We now almost understand why Russell insisted that ordinary names could not abbreviate purely qualitative definite descriptions, but necessarily abbreviate hybrid descriptions. I say "almost" because the epistemic constraint just mentioned does not yet fully explain why names cannot disguise purely qualitative descriptions. The epistemic constraint does entail that the name user be in possession of identifying knowledge, but this is compatible with the name abbreviating a description that is purely qualitative, a description that doesn't convey such identifying knowledge. Let the name user, for example, know that some appropriate hybrid description, for example, `the person causing this sense datum', applies to the denotation of the original purely qualitative description, for example, `the longest living person'.

The mere presence of an identifying knowledge requirement for ordinary names, then, while it seems related to Russell's insistence on the hybrid description picture, does not, by itself do the trick. The missing link--what is needed to see why names could not abbreviate purely qualitative descriptions--is provided by what I take to be Russell's intuition that not only do names require identifying knowledge for their use, but the names themselves indeed capture or convey that knowledge. They do this by abbreviating hybrid descriptions that formulate this knowledge. What names do for us--and this is Russell's "datum," the epistemological intuition that fuels the insistence on the hybrid descriptions model--is to identify their bearers or denotations, to indicate which things are in question. 

Russell, in motivating his idea that ordinary names conceal hybrid descriptions, misspoke. His point was not, if I understand him, that there is any sort of knowledge that one cannot have concerning the denotation of a purely qualitative description. His point was rather that the use, or understanding, of such a description--as opposed to the use, or understanding, of an ordinary name--does not capture or convey or express the sort of knowledge in question.

Russell's views about ordinary proper names seem to me underexplored. I've been focused upon the epistemic motivation for Russell's hybrid-descriptions account, but there is still more. There is in "Knowledge By Acquaintance and Knowledge By Description" a distinct, semantic motivation for his view of ordinary names. Consider these two passages.

Suppose some statement is made about Bismarck.  Assuming that there is such a thing as direct acquaintance with oneself, Bismarck might have used his name directly to designate the particular person with whom he was acquainted.  In this case, if he made a judgment about himself, he himself might be a constituent of the judgment.  Here the proper name 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. [Italics added.] But if a person who knew Bismarck made a judgment about him, the case is different.  [In this case since the speaker is not acquainted with Bismarck, the name must stand for a description.] ("Knowledge By Acquaintance and Knowledge By Description," p. 209.)

It would seem that, when we make a statement about something only known by description, we often intend to make our statement, not in the form involving the description, but about the actual thing described.  That is to say, when we say anything about Bismarck, we should like, if we could, to make the judgment which Bismarck alone can make, namely, the judgment of which he himself is a constituent.  In this we are necessarily defeated, since the actual Bismarck is unknown to us. ("Knowledge By Acquaintance and Knowledge By Description," p.210)

Russell's remarks are confusing.  In the first he tells us that all names, even ordinary ones, want, so to speak, to reach out and touch someone.  What does this mean? 

Russell suggests, in the second passage quoted, that the use of an ordinary proper names involves a frustrated referential intention. Surely, however, Russell is not speaking seriously here about referential intentions and their failures analogous, say, to the remark that the ancient Greeks intended to refer to Zeus, but were frustrated in this intention due to Zeus's lack of existence.  Imagine that I really intend to use an ordinary name as a genuine name, that is, I do not mean to put forth a descriptive characterization of the object of reference.  If I am not, however, in an epistemic position to directly refer, then my intended reference should turn out to be a real reference failure, not a successful use of a device that abbreviates a definite description.  What then is Russell's point?

Perhaps Russell has in mind, in this second passage--the first remains mysterious--not referential intentions, but rather referential wishes.  Perhaps his point is that we would like to directly refer to Bismarck, which is not to say that we utter the name with the intention of directly referring to him.  As much as we would like to so refer to him, we cannot and so we do what is second best and express a descriptive characterization of him.  How coherent is this?  I'm not sure.  It seems to unrealistically assume great semantic sophistication on the part of ordinary speakers:  they would like to refer directly, but since (they know that?) this is impossible, they go in for the descriptive mode of expression.  Referential wishes do not seem to help here, any more than did referential intentions.

I want to suggest a different interpretive direction. Russell gives voice in both of these passages to a lingering sense that somehow names, ordinary ones, do not quite feel descriptional. Isn't it just this sense that Russell attempts to evoke with the remark that 'Bismarck', as Bismarck himself uses it, has the direct use that it always wishes to have, or with the comment that even when I use 'Bismarck', the name tries to directly get through to the man himself?  If not for what we know about the epistemic conditions for real naming, we can imagine Russell musing, it would be tempting to treat all names, ordinary ones included, as directly referential.  Russell, if I am not mistaken, felt a conflict between the dictates of his semantic ear,⁽¹⁸⁾ according to which names are directly referential, and his epistemological conscience.

I see in Russell's discussions of ordinary names, then, a tension between two conflicting pictures of the semantics of ordinary names. Russell's dominant tendency, the one that gets all the press, is to treat ordinary names as disguised definite descriptions. One may begin to get the scent of a conflicting tendency when Russell so often uses ordinary proper names as examples of genuine names, and characteristically remarks in such contexts that he is speaking as if mere ordinary names were real names.  Perhaps, but Russell's direct reference impulse shines through most clearly, I think, in the quoted passages.

What we might call "Russell's compromise," his idea that ordinary names are descriptional-but-not-purely-so, might well represent, at least in part, his attempt to do justice to the conflict between his semantic ear and his epistemological scruples. Ordinary names fail to make direct semantic contact with their referents--as dictated by epistemology.  Yet they do not merely talk about things by indirection, as do purely qualitative definite descriptions.  Ordinary names, since they abbreviate definite descriptions that contain genuine names of particulars, make a kind of semi-direct contact with their bearers by specifying the latter in terms of directly referred-to particulars.

I have explored two philosophical motives for Russell's hybrid description theory of ordinary names, epistemic and semantic.  Russell's ideas here are, as usual, ingenious and intriguing, as well as very frustrating.  Russell is a source of great insight and suggestion, but also of ideas that are, at best, half worked out.  I find his epistemological ideas especially frustrating. Perhaps this is because they presuppose a Cartesian-inspired epistemology that does not recommend itself to me.  But it is also because of Russell's sloppiness, the easy use he makes, for example, of terribly obscure notions like knowing who.

One cannot deny, however, the richness of Russell's epistemic ideas, for example the idea that there are epistemic requirements for the use of even ordinary proper names, that the use of a name requires some kind of identifying knowledge.  This idea is independent of, although of course not unrelated to, the thesis that names are description-surrogates. The epistemic requirement thesis, like so many of Russell's ideas, shows up, albeit in different dress, in contemporary discussions.⁽¹⁹⁾ 

It is striking how differently Frege and Russell respond to the challenges names present. Russell, unlike Frege, was impressed, almost obsessed, with proper names, even ordinary ones. Ordinary names, as we have seen, constituted a very special category to Russell. Indeed, if my reading of Russell is on track, ordinary names--both semantically and epistemologically--are much more like genuine names than the usual talk of the "Frege-Russell description theory" would suggest. Frege, by stark contrast, introduces a category of "proper name" with the remark,

The designation of a single object can also consist of several words or other signs. For brevity, let every such designation be called a proper name. ("On Sense and Reference," p.57)

Frege is thus not all that worried about what is distinctive about actual proper names. Semantically, they are of a piece with definite descriptions.

It is unfortunate that Russell never really develops, or even makes fully explicit, his semantic motivation for the hybrid descriptions view of ordinary names, what I have referred to as the dictates of his semantic ear.  Here, in a way that is entirely independent of his Cartesian epistemology, Russell's deepest differences with Frege are highlighted.  And here, Russell anticipates a most influential idea in contemporary philosophy of language, the idea that an ordinary name is a "mere noise or shape conventionally used to designate a certain [thing]; it gives us no information about that [thing], and has nothing that can be called its meaning as opposed to denotation."

Despite the deep differences uncovered between Russell and Frege, there remains, at least so far, this substantial agreement:  ordinary singular terms like names and indexical expressions refer by expressing concepts. Well, perhaps "concepts" are not quite right for Russell. Ordinary singular terms, after all, express constellations of universals-cum-(acquainted-with) particulars, and referring to these as concepts is perhaps distorting Russell's view, excessively assimilating it to Frege's.⁽²⁰⁾ Still, Russell apparently shares with Frege the idea that ordinary singular terms make mediated contact with their referents. The mediating agents may be Fregean senses, or Russellian universal-cum-particular constellations, but they mediate nevertheless. Or so it seems. The problem is that Russell's famous theory of descriptions suggests otherwise, for according to it, definite descriptions do not really express constellations of universals and particulars, nor do they even really refer to anything.⁽²¹⁾  Let's see why. 

One thing that is novel about Russell's approach to the semantics of definite descriptions, and particularly relevant here, is Russell's idea that a sentence of the form "The F is G" is, to dramatize a bit, almost ill-formed.  Less dramatically, the grammatical structure of such a sentence, its subject- predicate form, gives a misleading picture of the content of the sentence, of what the sentence is actually putting forth as true.  Such a sentence does not, contrary to the grammatical appearances, attribute a property, G, to a thing, the F, as do sentences of the form "a is G," where `a' is a logically proper name.  How, then, are we to understand such sentences? What exactly do they assert?

Russell's well-known answer⁽²²⁾ is that the content of such sentences is perspicuously formulated thus:  One and only one thing has F, and that very thing also has G; or, more long-windedly, something has F, anything that has F is identical to this first thing (so that no more than one thing has F), and that very thing also has G.  One who says, "The President of the United States who emancipated the slaves was a Republican," should not be thought of as, strictly speaking, attributing the property of being a Republican to Lincoln.  Don't think of this utterance as mentioning Lincoln at all.  Think of it as saying something very general about the universe:  There is one and only one thing that has a certain property, being a United States president who emancipated the slaves, and, further, this very thing has an additional property, being a Republican.

While the general direction of Russell's approach is clear, the details are quite another matter.  For one thing, the example just given, like many of the examples that spring to mind, including many of Russell's examples, would be extremely complicated to work out in detail.  We might symbolically represent the reformulation just discussed as

(Ex) {(Px) & (y) [Py --> (y = x)] & Rx}

What property does `P' designate?  Remember, the predicate `is a United States President who emancipated the slaves' embeds another description, `the slaves', as well as the name `United States'. The latter name in turn abbreviates a hybrid description, according to Russell.  So we should think of the above reformulation as a mere first step towards a final version in which all descriptions are eliminated.

Even more troublesome is obtaining a precise account of the semantics of Russell's perspicuous formulations.  This would involve, among other things, an account of the semantics of quantifiers, and Russell himself talks about this question in confusing, and on different occasions quite different, ways.  Who knows how all of this is to ultimately get spelled out?

Let's be content, then, with Russell's basic idea, that the old, description-containing sentence ought to be scrapped in favor of one that employs quantifiers, variables, and so on, so as to perspicuously formulate the proposition.

I want to focus upon the great gap between the look of the original sentence and that of the reformulation.  A radical metamorphosis seems to have transpired.  The definite description does not simply get replaced by some other expression that better exhibits the description's function. We might think of Frege's view in this latter, less radical, way: A name, according to Frege, fails to make explicit its informational content, and so were we to replace it by a definite description, this would have the virtue of making the content explicit. This latter kind of reformulation leaves the basic structure intact. 

Russell's reformulation, by contrast, yields a radically new structure. What looked like (but never really was) a reference to Lincoln is gone, and instead we have references to various properties, acquainted-with-particulars, and so on (including whatever references are involved in the use of quantifiers, possibly propositional functions, second-order properties, and who-knows-what).  The result is a new sentence that, as a whole, is supposed to be functionally equivalent⁽²³⁾ to the original, but not at all, so to speak, piece by piece.  The basic structure of the propositions we express by ordinary sentences is, according to Russell, radically different from the structure suggested by the grammar of those sentences, and radically different from what, Russell admits, one would have naturally assumed.⁽²⁴⁾,⁽²⁵⁾

Consider now a Russellian reformulator, one who takes his Russell very seriously.  (There is one in every crowd.)  The Russellian reformulator looks at definite descriptions as mere artifacts of the misleading grammar of ordinary language.  Definite descriptions don't really exist for him, at least not at the level of, as he likes to say, logical form.  When a Russellian reformulator hears someone utter a sentence that contains one or more definite descriptions (or description abbreviations), he barely hears the vulgar form(s) of expression. What registers is "what the speaker is `really' saying," the perspicuous Russellian reformulation. (Think of yourself "translating" a child's poorly constructed remarks.)

Does it make any sense, given this Russellian perspective, to speak of "the constellation of universals-cum-particulars expressed by a definite description"?  If we took the unanalyzed sentence at face value, if we left the grammatically unified definite description intact, it would have been natural to speak of "what it expresses," a concept, constellation, or whatever. We are, however, not leaving this expression intact. When we straighten out what is being said, there is no unified expression that plays the role of the description. The Russellian reformulation eliminates the description, and disburses those expressions that derive from it. There is thus no expression left that can be thought of as expressing such a constellation. Nor, to speak more ontologically, is there any such constellation of universals-cum-particulars that enters as a unit into the proposition.

Talk of names and descriptions expressing semantically mediating entities, while it fits hand-in-glove with Frege's approach, it does not seem felicitous with respect to Russell's perspicuous reformulations.  Strictly and philosophically speaking, then, Russell does not think of descriptions, nor of the ordinary names and indexicals that abbreviate descriptions, as expressing constellations of concepts and particulars. What has become of what we took to be Russell's Fregean tendency, his representationalism?

It will be instructive here, before we get carried away with strict and philosophical talk, to take brief note of Russell's introduction of the notion of "denotation."  Russell's maintains that although it's no part of the linguistic function of a definite description to pick something out (how could it be since these expressions don't exist at the level of logical form?), we can still speak of the "denotation" of a description.  The perspicuous reformulation of "The President who emancipated the slaves was a Republican," may not contain any expression that refers to Lincoln, but there is still a sense in which it is "about" Lincoln.  That sense emerges when we note that there is a unique individual particularly relevant to the truth or falsity of the perspicuous formulation. Let us, suggests Russell, call this individual, if indeed there exists such, "the denotation." 

Thus if `C' is a denoting phrase [e.g. a definite description], it may happen that there is one entity x (there cannot be more than one) for which the proposition `x is identical with C' is true, this proposition being interpreted as above. [Take the description `the F', for example, in the context of the sentence `x is identical to the F', and eliminate the description in accordance with the theory of descriptions, thus perspicuously reformulating the proposition.] We may then say that the entity x is the denotation of the phrase `C'. ("On Denoting," p.51)

Definite descriptions may not really function to pick something out, but it is harmless enough to speak as if they did, since, in the felicitous case at least, there will be a unique individual particularly relevant to the truth or falsity of the perspicuous formulation.

Russell's "philosophy of `as if'" approach to "denotation" seems applicable as well to talk of the concept or constellation expressed by a definite description. The Russellian reformulator, although he may balk at talk of definite descriptions expressing concepts, is not altogether ignorant of what we might reasonably call "the meaning" of individual descriptions (even considered "in isolation," perish the thought).  For any definite description, there will be universals, acquainted-with-particulars, and so on, that will get mentioned, albeit in a disbursed fashion, in the perspicuous reformulation. We can, if we like, mentally collect these disbursed remains of a description, and speak as if there were a unified entity expressed. Such talk, properly understood, is harmless, and it facilitates recognition of what still can be called the "representationalist character" of Russell's approach.  Just as the description can be said to be about someone, in the sense of having a denotation, it can be said to be about that individual in virtue of the individual's fitting the reconstituted constellation.

Russell's representationalism, it turns out, is a trickier business than Frege's, trickier than we had supposed.  Indeed, it is, in one important respect, a more severe representationalism than is Frege's. Not only can one not speak of, for example, external things, (more generally, anything that is not an object of acquaintance) in a conceptually unmediated way.  One can never really mention such things at all.  Such an epistemically removed item can never be the subject of a subject-predicate proposition, at least not one that we can understand.  One can "talk about" such things only in the oblique fashion of Russell's perspicuous reformulations. We can, in this oblique sense, "talk about" Lincoln by asserting that: one and only one thing was a presidential slave emancipator, and that thing was a Republican.

Frege and Russell seem so far apart on so many of the questions at which we have been looking that it becomes difficult to discern anything that deserves to be called a Frege-Russell approach to semantics. They are worlds apart on such fundamental topics as the nature of propositions, the intelligibility of direct reference, the viability of the sense-reference distinction, and the integrity--as we might call it--of the syntax of natural language. Even where they seem to agree, moreover, one needs to take a closer look. They seem to agree, for example, that ordinary proper names have descriptive content. What this comes to, though, is very different for Frege, than it is for Russell.

It may be useful, nevertheless, to revert to a level of abstraction at which they do agree. If, as many argue nowadays, ordinary proper names fail to have anything like the close relation to definite descriptions that both Frege and Russell, each in his own way, suppose, then descriptivism, in both its Fregean and Russellian embodiments, is incorrect. Nor is descriptivism, contrary to what is usually supposed, their most significant area of agreement. Their deepest, most fundamental, point of contact, as I see things, concerns what I will call the "intentionality intuition." This is the traditional idea that if one is to speak or think about a thing, one must possess a discriminating cognitive fix on the thing, that something about one's cognitive state must distinguish the relevant item from everything else in the universe. Otherwise, so the intuition goes, what would make this thing the referent?⁽²⁶⁾

Even here, however, Russell and Frege approach the matter in very different ways. They disagree on what counts as the appropriate sort of cognitive fix, on just what sort of cognitive relation is required between thinker and object of thought. Frege's idea was that reference to an object required that the object be brought under an individuating concept, that only the possession of an individuating concept puts one in a position to refer to that which satisfies the concept.

Russell was, by comparison with Frege, a fanatic about intentionality. Russell was most unsatisfied with the idea that the mere possession of a concept might supply the required cognitive grip. One might possess a purely qualitative concept that in fact applies to a certain entity, we can imagine Russell thinking, and yet have absolutely no idea who or what satisfies that concept. Were one to be directly acquainted with an object on the other hand, were one to have the object, as it were, smack up against one's mind, one would really know which thing was in question.

The Frege-Russell controversy about direct reference, then, takes place against the shared conception that reference to an object requires a discriminating cognitive fix. Frege's approach to the cognitive fix requirement makes direct, conceptually unmediated, reference an impossibility. Russell, given his Cartesian epistemology with its conception of epistemic intimacy, thinks that direct reference is indeed possible, even if only reference to epistemically select items.

I stated above that the most fundamental point of contact between Frege and Russell is not, to my thinking, their agreement on the descriptive character of names. Russell, for one thing, does not espouse a description theory for what he took to be the real names, a fact that is too often noted but quickly set aside in our enthusiasm for ( or against) their shared descriptivism. While Russell's descriptivism is thus mitigated, his view about the necessity of a strong cognitive fix, especially for genuine names, is not. Russell's tendency, moreover, was not to assimilate names to descriptions--as was Frege's--but to emphasize the differences. Even ordinary names, Russell held, involve a kind of epistemic intimacy with their referents not characteristic of purely qualitative descriptions, or so I've argued. If I'm correct, moreover, about Russell's semantic ear, he would have loved to find a way to view ordinary names as directly referential, if only their epistemology would have taken care of itself.

Nor is it even clear that descriptivism is at the heart of Frege's approach. Indeed, there is a controversy in the literature about whether Frege was a descriptivist at all, whether Frege took the senses of names to be given by purely qualitative definite descriptions. What is uncontroversial is that Frege took modes of presentation to be essential to reference. While it is difficult to imagine Frege as movable on the latter question, it is much easier to see the description theory as negotiable. For both Frege and Russell, then, it is not the description theory, but the cognitive fix requirement, that goes deepest.

Our quest has been for a common perspective shared by Frege and Russell. After locating many areas of disagreement, we have located a fundamental point of agreement, indeed, one that Frege and Russell share with many philosophers, past and present, one that may indeed look unassailable. I have underscored it, rather than descriptivism, because I believe that it furnishes, or ought to furnish, a more important target for contemporary critics of Frege and Russell. It's radical rejection, I want to suggest, is the deep lesson Putnam sloganized as "Meaning ain't in the head."

ENDNOTES

1. "Letter to Russell," Gottlob Frege: Philosophical and Mathematical Correspondence, Abridged Edition (University of Chicago Press, 1980), p. 169.

2. Talk of Russell's perspective (in semantics or elsewhere) is risky.  There were, of course, various strands to Russell's thinking, and these, moreover, took on different emphases at different times.  Nor did he hesitate to change his mind.  I will isolate one important strand in Russell's thinking, a strand that contrasts sharply with Frege's approach, and one that is, at least to my mind, suggestive of a way forward.  Russell emphasizes these views in "Knowledge By Acquaintance and Knowledge By Description," in Bertrand Russell, Mysticism and Logic, as well as in the parallel piece by the same name in The Problems of Philosophy, (Home University Library, 1912, reprinted 1959, Oxford University Press).

3. The terminology of "direct reference" derives so far as I know, from David Kaplan.  He uses it, however, in a wider way than I do here. As I use the terminology, a directly referential expression will be one that, as Russell says, is a "mere noise or shape conventionally used to designate a certain [thing]; it gives us no information about that [thing], and has nothing that can be called its meaning as opposed to denotation." Kaplan's notion, on the contrary, allows for directly referential expressions that possess descriptive meaning, for example, indexicals on Kaplan's view. 

4. In Chapter 1 of my forthcoming--as yet untitled--book, I argue, following Richard Mendelsohn in "Frege's Begriffsschrift Theory of Identity," Journal of the History of Philosophy, Vol.22, No. 3, July 1982, PP. 279-299, that modes of presentation were not the innovation of Frege's later sense-reference picture. What was new in "On Sense and Reference" was the idea that modes of presentation, already available to Frege in the earlier work, might well be seen as propositional constituents, and that this would make for a more natural solution to the puzzle of informative identity statements than had been available. See Section 8 of the Begriffsschrift for the earlier view.

5. What I am envisaging here is the possibility that ordinary names could function semantically as mere tags for their referents, a position that, as we will see, Mill maintained.  'Hesperus' and 'Phosphorus' would accordingly be tags for a single planet.  Russell, of course, would have rejected this idea for epistemological reasons, namely that we are not directly acquainted with the heavenly body.  The question here under consideration is why does Frege reject the idea.

6. For maximum plausibility, let's count any "way of thinking" as a concept.  Let's say that such cases count as "bringing the object under a concept."  I have considerable doubts as to whether Frege would so proceed, but others so interpret Frege, and I don't want anything to hinge on this question of interpretation.  If Frege is interpreted more strictly, so that 'concept' means, e.g.. "the sort of thing expressible by a definite description," then the conceptual representationalist thesis will seem more questionable.  But more of the latter later.

7. His reasons are another matter.  Russell's obscure discussion of his objection to the sense-reference approach, in "On Denoting," is not a paradigm of philosophy. 

8. Both descriptions derive from Russell. See "Knowledge By Acquaintance And Knowledge By Description."

9. Nathan Salmon, in Frege's Puzzle (MIT Press, 1986), p.64ff., speaks of such descriptions as "thoroughly descriptional," as opposed to those that are "relationally descriptional."

10. Russell is not here singling out place names for special treatment. The same holds for proper names like 'Bismarck', as the context makes clear.

11. I'm not sure why Russell, of all people, says that this description "must apply to some man." If two men share the honor of being longer lived than anyone else, so that there is no unique most long lived man, then the description lacks a denotation, as Russell's famous theory of definite descriptions reminds us. But let's not worry about this.  Russell seems willing to allow, in any case, that we can come to know that a description like 'the most long-lived of men' has a denotation.

12. The identification with knowing who is suggested by much of Russell's discussion in "Knowledge By Acquaintance And Knowledge By Description", and his parallel piece in The Problems of Philosophy. See, for example, Russell's discussion of his idea that even one who knows Bismarck only through history--and this is not, for Russell a case of having a mere purely qualitative description that denotes Bismarck--knows who he was. (The Problems of Philosophy, p.58.)

13. Russell sometimes appears to deny that in cases like that of my son's name, I can know who is in question.  He says, e.g., that only when I am acquainted with something can I know which thing is in question.  At the same time, he clearly holds that we can know who is in question in cases in which knowledge by acquaintance is out of the question, for example, the Bismarck case mentioned in the last note. Perhaps the point is that there is a strong sense of "knowing who," according to which one knows who x is only if one is acquainted with x.  But in another perfectly legitimate sense one can know who someone is if one knows enough about the individual, even if one is not acquainted with him/her.  See note 21 and the accompanying text for further discussion of "knowing who." The topic of "knowing who" is extremely complicated, as recent philosophy testifies, and no serious attempt at resolving the perplexities involved can be carried out here. For a recent illuminating discussion, the general tone of which I feel sympathy for, see Steven E. Boer and William G. Lycan, "Knowing Who," Philosophical Studies 28 (1975): 299-344.

14. knowing who, or perhaps Russell has something else in mind.

15. In the "weaker" sense of "knowing who" distinguished in note 16 above.

16. Here I adapt a remark of Russell ("Knowledge by Acquaintance and Knowledge by Description," p.210): 

If however, we say, "the first Chancellor of the German Empire was an astute diplomatist," we can only be assured of the truth of our judgment in virtue of something with which we are acquainted --usually a testimony heard or read.

17. Actually, there are varying strengths of epistemic access afforded by constellations of universals-cum-(acquainted-with-) particulars.  The constellation expressed by 'my daughter', after all, should be quite different in this respect than that expressed by `the individual, whoever that may be, who is standing closest to a point exactly 4000 miles due west of me at the moment'.  Russell would surely agree--he himself distinguishes in the passage quoted above between the "stage in the removal from acquaintance" represented by Bismarck's friends, and that represented by one who knows Bismarck only through history. Russell, then, would presumably not want to say (or at least he might well not want to say) that in cases like `the individual, whoever that may be, who is standing closest to a point exactly 4000 miles due west . . .', despite the agent's possession of a "hybrid concept," the agent automatically knows who is in question. So possession of a hybrid concept may not supply a sufficient condition for knowing who, and Russell never even hints at what further conditions might be relevant. In what follows I will ignore this complication, and speak as if the possession of any hybrid description assures the appropriate sort of identifying knowledge of the denotation.

18. The expression is Joseph Almog's.  Almog has argued in conversation that there are direct semantic intuitions to the effect that ordinary proper names are directly referential.  I find this appeal to intuition difficult to evaluate.

19. This is true even of contemporary philosophers who explicitly deny that ordinary names are surrogates for descriptions. Keith Donnellan, e.g., a defender of a direct reference view of ordinary names, nevertheless argues that one must stand in a privileged epistemic relation to an individual in order to refer to it by name.  See, e.g. his "Rigid Designators and the Contingent A Priori," in P. French, T. Uehling, and H. Wettstein, eds., Contemporary Perspectives in the Philosophy of Language (Minneapolis, 1979). 

20. At issue here is not Russell's, nor Frege's, use of the term 'concept' (or 'Begriff'). I am using the term concept as I have all along, to refer to the something like Fregean senses (as opposed to the things Frege calls "concepts", i.e. the references of predicates, or that Russell does, i.e. universals). My point is that it is stretching things to attribute to Russell the view that hybrid descriptions express such concepts.

21. I haven't forgotten Russell's view that they definite descriptions have "denotations." I return to this below.

22. See "On Denoting" for Russell's reasons for denying that the apparent grammatical structure reflects what is really going on, as well as his reasons for the positive view sketched here.

23. There is more than one way to understand the relation between the original sentence and the Russellian reformulation.  I have been taking this in the way that I think was Russell's dominant trend.  Russell, that is, tended to view the reformulation as making explicit the thought in the mind of the speaker, a thought that is misleadingly put by the original sentence.  One might do it differently (in the spirit of Quine, for example) and see the original as muddled (who knows what ordinary folk think when they use descriptions, and who cares?) and the reformulation as coming as close to the original as possible while meeting serious standards of acceptability, intelligibility, etc.  For more of the latter see Quine's Word and Object, esp. "The Ordered Pair as a Philosophical Paradigm."  Also see Russell's remarks in this Quinean direction in his "Mr. Strawson on Referring," in My Philosophical Development (Allen and Unwin, 1959), Chapter 18, Part 3.

24. At one point in "On Denoting," after formulating how things should go according to his own theory, Russell says, "This may seem a somewhat incredible interpretation; but I am not at present giving reasons, I am merely stating the theory." (P.44)

25. There is an interesting methodological contrast here with Frege.  As Kaplan notes in "What is Russell's Theory of Descriptions?," in Bertrand Russell: A Collection of Critical Essays, ed. D.F. Pears (Anchor Books, 1972), pp.227-244:

Russell and Frege were both interested in removing the logical imperfections of ordinary language but their methods were quite different. . . Where grammar called for entities whose nature was obscure, Frege attempted constructions, as with numbers, or a theory about the purported entities, as with propositions.  Thus he sought to preserve the integrity of ordinary language with ontological ingenuity.  Russell's response, at least in the case of definite descriptions, was by grammatical reconstrual and replacement.

26. As I am using `intentionality', it has nothing special to do with the problem, associated with Brentano, of how it is possible to think about something that does not exist. `Intentionality' here concerns the more general phenomenon of the "aboutness" of thought. The traditional picture to which Frege, Russell, and many others subscribe has it that thought can be about something only if that something is intellectually discriminable by the thinker.