In forming their theories for understanding propositions, both Gottlob Frege and Bertrand Russell were forced to address basic questions regarding the role of names.  Do names serve only as placeholders to denote certain entities, or is there something logically significant about exactly how the names pick out the entities with which they are associated — about their mode of presentation?  The names “The present Pope” and “John Paul II” seem to have a difference that goes beyond their different grammatical form; yet that difference cannot be thought to be so stark as to prevent them from denoting the same object.  How is it that the specific choice of name can affect one’s understanding of the proposition?

To Frege, the mode of presentation was certainly a significant aspect; in fact, he gave the mode of presentation a fundamental place in his ontology as the “sense” of the name, distinct both from its denotation and from the subjective ideas associated with the word.[1]  Russell, however, rejected the concept of sense and placed far less emphasis on what I will call the “intensional” aspects of names — the properties that they do not share with other names that denote the same entity.[2]  Instead, Russell chose to analyze the proposition containing the name, eliminating (or, perhaps making explicit) the intensional aspects of names through a process that showed radical differences between grammatical and logical form.  Yet Russell’s success in analyzing away the intensional aspects of names is anything but clear; the possibility of informative identities among sense-data or sensory universals poses a number of severe difficulties to his theory, and even if the theory is consistent, Russell’s process of analysis seems to remove all useful methods for the communication of thoughts.

            Intensional aspects can become relevant in distinguishing among names in two ways.  Since there is no danger of confusing names that denote different objects, intensional aspects are necessary only to resolve questions where two names denote the same object or where both names fail to denote any object at all.  Concerning the latter, if the means by which a name indicates its object is irrelevant, then all vacuous names — expressions whose grammatical form makes it seem that they denote an object, yet in fact fail to do so — must be, in some sense, equivalent.  Examples include proper names such as “Odysseus” or definite descriptions like “the present king of France”; they also include pieces of pure nonsense such as “Hrzbk.”  However, although both types of names are equally vacuous, the expressions of the former category seem somehow more meaningful than those of the latter. A robust logical theory should give content to this distinction and should also allow us to decide the status of propositions containing such expressions, such as “the king of France is bald” — a proposition which seems to generate contradictions if it held to be either true or false.

If we are to distinguish between “the present king of France” and “the present zorg of Blaxon,” something intensional that relates to the expressions themselves and not the objects they denote must be involved in their interpretation.  For Frege, this dilemma is easily resolved by the availability of sense; pure nonsense lacks sense, while “Odysseus” or “the present king of France” remain meaningful because they have sense even without denotation.  Thus, a proposition such as “the king of France is bald” takes no truth-value, since part of that expression is bedeutungslos and thus prohibits the whole from having a denotation (i.e., a truth-value).[3]  Russell, however, who does not allow for sense or for other intensional aspects, would find Frege’s method inapplicable to either problem; thus propositions containing seemingly meaningful descriptions and those containing gibberish would seem to be indistinguishable.[4]

The occurrence of vacuous proper names in propositions therefore seems to be fatal — and so, in the essay “On Denoting,”[5] Russell describes a process of whereby proper names and descriptions are, to the fullest extent possible, removed.  Indefinite descriptions are analyzed according to Frege’s quantificational logic;[6] in the case of definite descriptions, a contextual definition is offered that transforms “Scott is the author of Waverley” to “there is an x such that x wrote Waverley and no one else did, and furthermore x was Scott,” a form in which nothing corresponding to the original description appears.  In general, statements of the form “the F has G” become , a structure entirely without definite descriptions like “the F”; problems of vacuous descriptions can then be resolved using more familiar logical techniques. “The present king of France is bald” would become the sentence , which, because there is no such x, can be clearly determined to be false.  To resolve the status of sentences such as “Apollo is bald,” instances of “Apollo” can be replaced with a corresponding definite description — Russell suggests “the sun-god”[7]— and the sentence interpreted using previously outlined rules.  Of course, there is a need for iteration of this procedure if the definite description contains new proper names, and there must be a guarantee that the analysis will cease after a finite number of iterations, but these will be dealt with below.  What is important to note at this point is that as proper names and definite descriptions are eliminated, the danger of vacuity disappears, and the dilemmas noted above lose their relevance.

This process of analyzing sentences until the original names disappear is equally effective in resolving the two problems in which different names refer to the same objects: informative identity and “indirect speech.”[8]  Frege opens his discussion in “On Sense and Denotation” by examining the former.  He notes that identity must be a relation between names or signs rather than objects; one cannot say that two objects are identical, for then they would be the same object, and the statement “a = b” would seem to have no more content than “a = a.”  Yet statements of identity do seem to have content; statements like “The Morning Star = The Evening Star” are by no means a priori.[9]  If identity is a relation among names, however, it must be more than a merely formal relation among marks on paper; any symbol can be introduced to represent any other, and identity in such a view could provide only linguistic truths.  To cite the example from Frege’s letter to Philip Jourdain,[10] the statement “Afla = Ateb” does not tell us that “Ateb” is a symbol for “Afla,” but that the mountains observed and presented as Afla and Ateb are in fact the same mountain.[11]

How is informative identity then to be understood?  Frege can explain the difference in “cognitive value” through a difference in sense, as “Afla” and “Ateb” utilize different modes of presentation.[12]  Russell, however, has no such option, and must again eliminate the identities in the course of analysis.[13]  We have already seen what would happen to “Scott = the author of Waverley”; in the case of proper names, Afla and Ateb would be reduced to their descriptions “the F” and “the G,” which in Russell’s analysis would eventually reduce their identity to . The only identities here are between bound variables, where no mode of presentation is at issue and no issue of informative identity can arise.

The elimination of informative identity provides for the quick resolution of the problem of indirect speech on which Russell placed such emphasis.  Russell accepted the principle of indiscernability of identicals (also called Leibniz’s Law), that if a = b, then a may be substituted for b in any proposition without loss of truth.[14]  However, when such propositions involve indirect speech — the discussion of one proposition in another where it is “indifferent to the truth of the whole whether the subordinate clause is true or false” — this principle seems to break down.[15]  To cite Russell’s favorite example, the truth of “George IV wished to know whether Scott was the author of Waverley” is entirely separate from whether Scott indeed authored Waverley.  Yet Scott was the author of Waverley, which should by substitution should allow us to conclude that “George IV wished to know whether Scott was Scott,” which is false.[16]

The crucial point in this discussion is that for Russell, the problem of substitution in indirect speech appears only through the use of informative identities.[17]  Consider the statement “George IV believes A has H.”  If I am to substitute something into that sentence, I must be sure that an identity exists that justifies the substitution.  Furthermore, that identity may not be merely formal.  If I introduce the name “B” to symbolize A, so that “A = B” is stipulated as true, then “George IV believes B has H” will, given this symbolism, in fact be true and cause no contradiction — even if His Royal Majesty would never have expressed his belief in those exact terms.  To create the contradiction, it is necessary to impute a belief to George IV that he did not have, meaning that the substitution must contain additional knowledge; without informative identities, the premise of Leibniz’s Law will never be satisfied.  The statement said to be believed by George IV, namely that “A has H,” will often itself be changed by analysis, and Russell in “On Denoting” uses this fact, rather than the elimination of informative identity, to resolve the problems posed by indirect speech; yet this approach, though correct, is unnecessary.  As long as informative identities may be analyzed away, propositions involving indirect speech will not experience meaningful substitutions and will pose no danger to the theory of descriptions.

Russell thus seems secure as long as informative identities may be removed through analysis.  Yet Russell has such option where the informative identity “A = B” is already fully analyzed and need not be restated in a different logical form; should such an identity ever arise, there will be no escape from the contradictions.[18] The question of whether fully analyzed terms may be found in informative identities is not, to my knowledge, explicitly addressed by Russell; yet it is a question of fundamental importance to his theory.  If fully analyzed identities can be proven impossible, then the theory is well-protected from the dangers posed by intensional aspects of names; if such identities can be found to exist, then the theory may well be inconsistent.

To determine whether fully analyzed and informative identities exist, it is necessary to examine the meaning of “fully analyzed.”  As stated earlier, the process of analysis that Russell outlines in “On Denoting” requires, must be an iterated process; understanding “Aristotle” as “the student of Plato and teacher of Alexander”[19] would require descriptions of Plato and Alexander — and so on ad infinitum, unless there is a basic level at which the analysis ceases, the full logical structure of the proposition is revealed, and we may make our judgment as to its truth or falsity.  To create this basic level, Russell introduces the notion of acquaintance:  we are acquainted with certain entities, and those entities with which we are not acquainted can, after a sufficient amount of analysis, be expressed in descriptions composed only of entities with which we are acquainted.[20]  For brevity, I will speak of “acquainted entities” and “described entities”:  as is clear from the process of analysis, acquainted entities may be mentioned only through the use of proper names, while described entities must either begin as definite descriptions or be translated into them.[21]

In its general sense, Russell gives little explanation of the notion of acquaintance — among other things, he says one is acquainted with an object when one has a “direct cognitive relation” to it, but what cognitive relations may be considered “direct” and what objects might be termed acquainted depends on his epistemology rather than his logic.[22]  From logic’s perspective, acquaintance serves only as a formal distinction and a terminus in the analysis of sentences; the proper name of an acquainted entity must not be vacuous, and whether an entity is acquainted or described must always be decidable, but otherwise logic imposes no limits on the number or kind of acquainted objects.  If one’s epistemology allows for direct cognitive relations with angels, then angels may be considered acquainted entities and one may mention them in fully analyzed propositions.

            We have already seen that if an informative identity is to be found, it must come at the level of acquainted entities.  Yet if one chooses an epistemology that allows for knowledge, say, of angels, what would prevent an informative identity of the form “Angel A = Angel B”?  It would seem that Russell’s notion of acquaintance must, in order to ensure his consistency, give content to the notion of “direct cognitive relation” such that two informative names of acquainted entities are prevented from referring to the same object.[23]  Such a clarification is attempted in The Problems of Philosophy,[24] where he states that direct awareness of an entity is awareness “without the intermediary of any process of inference or any knowledge of truths.”[25]  Russell gives as an example a sense-datum of a certain patch of color; although many things may be said about it — “that it is brown, that it is rather dark” — those truths, or others that might later be learned, do not reveal anything in the sense-datum that was not known before.  His knowledge of the color is perfect and complete, “and no further knowledge of it itself is even theoretically possible”; acquainted entities are thus those “immediately known to me just as they are.”[26]

            This property of complete knowledge is what recommends protects Russell against informative identities, and his argument that knowledge of sense-data is complete recommends his empiricist epistemology in which only sense-data can be acquainted.[27]  Identities, to be informative, must contain information about the entity named that was not known before; “Afla = Ateb” would be trivial if all of the properties of the mountain were already assumed when one spoke of it.  To claim such complete knowledge for mountains or angels requires a rather bold epistemology; yet when dealing with a sense-datum, there seems to be a logical guarantee of complete knowledge.  My sense-data are such by virtue of the fact that I perceive them, and there cannot be any aspect of what I perceive that I have not yet perceived.  Although my perception of the letters on the eye doctor’s chart might become more sharp as I squint and pay attention to them, that does not deprive my earlier blurry perceptions of their status as having been my actual sense-data.[28]

            If this analysis is correct, then it would seem impossible to show informative identities among sense-data.  Yet Russell himself recognized that this doctrine of complete knowledge encounters “some difficulty” concerning the definition of “one sense-datum.”[29]  Russell notes that sense-data may be complex, and that further attention may reveal new information about the relations among their parts;[30] however, he gives no means of determining whether a complex sense-datum has been fully broken into constituents.  A patch of color may be understood as a large number of smaller patches, and thus be divisible in space; a given sense-datum may seem to persist for a certain duration, and thus be divisible in time; and if there are in fact “point-sized” sense-data sufficiently limited in time and space as to be indivisible, it would be near-impossible for Russell to state what the appropriate limits might be.[31]

While this line of approach seeks to restrict the scope of “one sense-datum,” the need to avoid informative identity makes one wish the scope expanded.  Daily experience offers us sense-data that are very similar, and it is possible that one might perceive two sense-data that seem, for all intents and purposes, exactly alike. They again might be separated in space, as in the case of two patches of white on a computer screen, or in time, as in the case of sense-data perceived before and after one blinks. If the perception of the patch of color that constitutes each of the two sense-data (call them A and B) is precisely identical, then the distinction between them — the reason “A = B” does not hold — must lie in the fact that they cannot be substituted for each other in propositions that discuss their spatial or temporal ordering, such as “A is before B.”[32]  However, such a sense-datum must not be just the particular patch of color that is perceived; it must contain information about its relations as well, which is quite unusual.  Given a sense-datum that is half-red and half-blue, one should be able to speak of a component blue sense-datum without also mentioning the red adjacent to it or the music that was playing in the background; one should be able to pick out particular elements and refer to them as individuals.  Yet it now seems that the relations of a sense-datum to its fellows must be essential and inextricable parts of its character; one may not speak of sense-datum A without including within it one’s perception (one’s sense-datum, it would seem) of A’s ordering relative to other sense-data in time and space.  This lack of independence seems problematic, especially since Russell describes the process of becoming acquainted with spatial and temporal relations as one of abstraction from pre-existing sense-data.[33]

            Such concerns may render the presence of informative identities among sense-data likely, but they do not conclusively establish their existence, relying as they do upon the ambiguity in the concept of “one sense-datum.”  Yet sense-data are not the only acquainted entities; according to Russell, we are also acquainted with universals, viewed by Russell as qualities that may be shared by many particulars.[34]  Russell charts the development of a universal as follows:  we begin with acquaintance with particular yellows, and “if we have seen a sufficient number of yellows and have sufficient intelligence, we are aware of the universal yellow.”[35]  Given that universals may be subject terms rather than only predicates, as in Frege,[36] there seems no reason why they may not be informatively identical — and there is good reason to think that some are.

If universals are abstracted from particulars, then any useful sensory universal must be general enough to include sense-data which we have not yet perceived.  For instance, we may experience sense-data of shades of color A, B, and C, and later abstract from them the universal “aquamarine” which means something more than “exactly resembles either A, B, or C.”  Thus other similarities between these three sense-data may be ignored in the process of forming a specific universal — for instance, whether they are mottled or smooth, or whether they would be better described as light aquamarine or dark — and there does not seem to be any clear limit on how general the abstracted similarities may be.  Looking at a group of objects from which I might abstract “aquamarine,” I could consider more general similarities and abstract the universal a.  I could follow the same process with a group of objects from which one might abstract “navy” and, at a higher level of generality, instead abstract the universal b.  It seems entirely possible that the universals I abstract separately from each might in fact be precisely identical and also identical to “blue” — implying, it seems, that “a = b” is a valid informative identity.[37]

            One might object that a and b ought not be considered identical but merely coextensive.  Russell does allow for distinct coextensive predicates — in fact, his Axiom of Reducibility stipulates them — but there seems no good reason to confine observations of similarity among sensory universals (all of which are of the same type and order) to coextensionality.  Universals are terms, and it should be possible to use them as participants in identity relations rather than only in predicative statements.  Any statement I make about a would seem to hold equally well of B; the only cases I am aware of in which the substitution fails would be those of indirect speech, since one may believe “a = a“ without holding that “a = b.”[38]  To prohibit this use of identity because it causes contradictions elsewhere in the theory is logically sound but highly unsatisfying; it smacks of “solving” Russell’s Paradox by stipulating that only those sets exist whose existence does not lead to a contradiction.  At the same time, however, this may not be considered a clear refutation of Russell’s views.  There is no guarantee that universals may in fact be informatively identical — and thus no conclusion as to whether any intensional aspects of names are required for one to understand them.

            Yet even assuming that Russell has been successful in eliminating from his system intensional aspects of names, his success has come at a not insignificant cost.  Frege’s concept of sense served not only to explain certain puzzles, but also to establish a public meaning for a word or proposition, something distinct from the subjective ideas a word may conjure.  In analyzing away all concept of sense, Russell also analyzed away all objects that could be perceived by more than one individual — all entities that could be described by Frege’s “common store of thoughts” shared by mankind.[39]  The subjects of fully analyzed propositions such as “this is yellow” may offer no mode of presentation, but as a result they make their truth-grounds for anyone but the speaker utterly obscure.  Constructions of sense-data are unintelligible to those who have not experienced the same sense-data — and the concept of “same sense-data” is empty if all informative identities have been removed.  Even if I have experienced the same “this” as the speaker of “this is yellow,” without informative identities of universals, I would be forever uncertain whether the “yellow” mentioned is in fact the universal I call by that name.[40]  Worst of all, if communication of this sort is prohibited, Russell’s theory itself may not stated in a way that enables it to be understood.  Although the problems of intensionality cannot show Russell’s system to be inconsistent, the concessions he makes to avoid contradiction leave the theory painfully limited and perhaps unsatisfying to those seeking a means of understanding propositions.

[1] I shall use “denotation” as a translation of Frege’s bedeutung.  The emphasis which Frege placed on sense is evident from his declaration that “for the purpose of acquiring knowledge,” the sense of a sentence was to be treated as “no less relevant than its [denotation], i.e. its truth-value” (“On Sinn and Bedeutung,” 1892, The Frege Reader, ed. Michael Beaney  (Oxford:  Blackwell Publishers, 1997), p. 171).

[2] I use this term with the strong suspicion that my usage is incorrect, but its use here seems analogous to its use in specifying the differences between coextensive concepts; in any case, this understanding is employed merely to supply a label and not to interpret the use of the word in other texts.

[3] Of course, the ease with which Frege answers the present question comes at the cost of much difficulty elsewhere, namely how one is to know that such statements as “Kepler died in misery” in fact take a truth-value. (“On Sinn and Bedeutung,” p. 162) His conclusion is that the use of such names in a logical statement presupposes that they denote something, and that such names would be eliminated through the proper construction of a “logically perfect language” (“On Sinn and Bedeutung,” p. 163).  A further difficulty concerns the exact sense of some proper names; while “the present king of France” has a clear mode of presentation, “Aristotle” does not, and Frege is forced to accept that different people may use the term with a different sense as long as the denotation remains the same — another issue that “ought not to occur in a perfect language” (“On Sinn and Bedeutung,” p. 153).

[4] Russell also may not accept the truth-value gap; since he views propositions as actual objects rather than (as Frege does) names for The True or The False, he must instantiate laws of logic (such as ) with propositions rather than truth values, and a proposition neither true nor false would therefore violate the law of the excluded middle.

[5] Bertrand Russell, “On Denoting,” Mind n.s. 14 (October 1905), pp. 479-93.  Page citations will be taken from the edition photocopied and handed out in class.

[6] E.g., “all men are mortal” becomes .

[7] “On Denoting,” p. 117.

[8] “On Sinn and Bedeutung,” p. 159.

[9] “On Sinn and Bedeutung,” p. 151.

[10] Letter to Phillip Jourdain, Jan. 1914, The Frege Reader, pp. 319-322.

[11] Letter to Jourdain, p. 320-1.

[12] “On Sinn and Bedeutung,” p. 152.

[13] Cf. Bertrand Russell, “Knowledge by Acquaintance and Knowledge by Description,” 1910-11, Mysticism and Logic (London: George Allen, 1970) pp. 163-4.

[14] “On Denoting,” p. 110.  The principle may be logically expressed as .

[15] “On Sinn and Bedeutung,” p. 160.

[16] “On Denoting,” p. 110.

[17] For Frege, the problem is somewhat more severe; since propositions normally denote their truth-values, it would seem that “George IV believes P” can be substituted for “George IV believes Q” as long as P and Q are either both true or both false.  Frege therefore introduces the theory that propositions used in indirect speech denote their usual sense rather than their truth-value (“On Sinn and Bedeutung,” p. 160).  Russell, however, does not believe that true propositions all denote the same object, and so may deal with indirect speech by addressing informative identities alone.

[18] In this case, “George believes that A has H” would imply that “George believes that B has H,” and although the use of “H” (and of “George”) in these propositions might be altered through analysis, if A and B are fully analyzed, the implication will remain valid.

[19] “On Sinn and Bedeutung,” p. 153.

[20] “On Denoting,” pp. 103-4; “Knowledge By Acquaintance,” pp. 158-9.

[21] This is because all definite descriptions clearly refer to described entities.  The class of proper names that Russell uses to denote acquainted entities is very small, however; in “Knowledge by Acquaintance,” he limits the class to “I” and “this,” and in a footnote added several years later even rejects the “I” (p. 162).

[22] The shift in terminology here from “acquainted entity” to “acquainted object” reflects Russell’s realism, in which acquainted entities such as relations and universals were considered objects that could serve as the subject of judgments.

[23] By “informative names,” I mean two names that offer different modes of presentation; a pair consisting of a name and a symbol stipulated to refer to the same object as the name would not be considered informative.

[24] Bertrand Russell, The Problems of Philosophy, 1912 (Oxford University Press, 1971), available online at http://www.ditext.com/russell/russell.html.

[25] The Problems of Philosophy, http://www.ditext.com/russell/rus10.html.

[26] Ibid.

[27] Russell also accepts logical forms and universals as acquainted objects (“Knowledge by Acquaintance,” p. 154); the latter will be addressed below.

[28] While this argument explains the exalted status afforded sense-data within Russell’s system, it also undercuts his efforts in “Relations of Sense-Data” to build a physics containing not only sense-data, but also “sensibilia,” the entities which are of the same nature as sense-data but are not acquainted.  Yet if we are not acquainted with sensibilia, whenever one is mentioned we must know it through a description composed of sense-data, which eliminates any explanatory advantage of a system comprised of both sense-data and sensibilia over one containing sense-data alone.  Although Russell would like to eventually eliminate sensibilia from his system and remove the “hypothetical scaffolding” (p. 117) — tossing away the ladder, as it were — it is not clear how this can be done while at the same time preserving the concept of an object persisting in time.

[29] “The Relation of Sense-Data to Physics,” Mysticism and Logic, p. 109.  Emphasis in original.

[30] “Relation of Sense-Data,” p. 109.

[31] For one thing, it would be absurd to say that we are acquainted with the limits unless we can give a very straightforward account of them; yet if they are known only through description, then they must be describable in terms of sense-data, which seems equally absurd.  Furthermore, the concept of sense-data as instantaneous would violate Russell’s principle that “no two instants are contiguous” (“On the Notion of Cause,” 1912, Mysticism and Logic, p. 135); it would also greatly undermine his project of representing time as similar in its continuity and completeness to the real number line.  It was this approach that enabled him to set the work of Weierstrass and Cantor against the paradoxes that were once thought to attend a time composed of instants (Cf. “Mathematics and the Metaphysicians,” 1901, Mysticism and Logic, pp. 63-70).

[32] One might object to this line of reasoning by saying that the names A and B contain no mode of presentation and offer no information, and therefore that no need exists to prohibit the proposition “A = B.”  This objection will be addressed below.

[33] The Problems of Philosophy, http://www.ditext.com/russell/rus10.html.

[34] The Problems of Philosophy, http://www.ditext.com/russell/rus9.html.

[35] “Knowledge by Acquaintance,” p. 154.

[36] Russell must be able to treat some predicates as subject terms in order to avoid the Kerry Paradox, which would make false any statement of the form “x is a predicate.”  While Frege’s defenders are able to claim that such statements are faults of ordinary speech to be avoided in a perfect language, Russell has no such excuse.

[37] These similarities need not occur only with the sense of sight; one might abstract the universal pineapple-taste from a few pieces of pineapple and the universal fruitsnack-taste from a few artificially-flavored fruit snacks and then later discover, upon further reflection, that the universals are the same; this is not an assertion that the fruit snacks taste like pineapple, but rather one about the taste universals themselves.  Because of their origins through separate abstractions, the universals compared here are not identical because of some mere formalism.

[38] I am assuming here that Russell would accept the biconditional form of Leibniz’s Law as a sufficient criterion of identity for universals.  Given that formal identities operate on this principle, no easily apparent reason exists why informative identities should not do so as well.

[39] “On Sinn and Bedeutung,” p. 154.  In suggesting that definite descriptions be assigned to each proper name, Russell encounters a similar difficulty to that which faced Frege in determining a proper name’s sense; specifically, Russell lacked a means of determining, for a given name, what the correct definite description would be.  At one point, Russell suggests that “Julius Caesar” should be understood as “the man whose name is Julius Caesar,” on the grounds that “whatever else I may have forgotten about him, it is plain that when I mention him I have not forgotten that that was his name” (“Knowledge by Acquaintance,” p. 160); he even claims that the statement “one and only one man was called Julius Caesar, and that one was assassinated” is fully analyzed and “wholly reduced to constituents with which we are acquainted” (“Knowledge by Acquaintance,” p. 161).  However, this view encounters several immediate dangers; one might speak of “the man whose name was George Sand,” say, or “the man whose name was Nicolas Bourbaki” (not even the pseudonym of one individual, but of several) without knowing that those descriptions are in fact vacuous. Furthermore, any clarification of the process of naming must involve some reference to those who bestowed the name on an individual; otherwise, one might interpret “Julius Caesar” to mean “the man whom I name Julius Caesar” — which seems entirely circular.  Yet without an objective route from proper names to descriptions, Russell provides no guarantee that statements employing proper names can be understood.

[40] There is also the possibility that the universals are coextensive; however, the finite limits of my experience of particulars falling under any two universals would give me no grounds on which to infer coextensionality.