Non-corrected new draft
Appendix to Chapter I
How Do Proper Names Really Work?
(Cutting the Gordian Knot)
Once fashion comes in, objectivity goes.
D. M. Armstrong
As Wittgenstein once said, our aim in teaching philosophy should not be to give people the food they like, but rather to offer them new and different food in order to improve their tastes. This is my intention here. I am firmly convinced that I have a much better explanation for the mechanisms of reference that characterize proper names, but the really difficult task seems to be convincing others. This difficulty is even greater, because I am swimming against the present mainstream – in this case, the externalist-causalist and often anti-cognitivist view regarding the meaning and reference of proper names.
There is a further reason why the neo-descriptivist theory of proper names that I intend to summarize here is particularly hard to accept. This is because the question of how proper names refer has always been the touchstone for theories of reference. More than forty years ago, when Saul Kripke, Keith Donnellan and others rejected descriptivism for proper names, they also opened the door for externalist, causalist and potentially non-cognitivist views concerning the reference of indexicals, natural kind terms and statements. Now, if I achieve my goal, which is to re-establish descriptivism concerning proper names in a more sophisticated way, the doors will again be open to re-establishing descriptivist-internalist-cognitivist views about other terms and language in general. This means that we will once again have to survey the whole topography of our philosophy of language. However, since the new orthodoxy is already well-entrenched – it has led a good life for the past forty years – and a myriad of good and bad arguments have been developed in its favor, the challenge is naturally great. If I limited myself to answering just the most relevant arguments, I would still need to write an entire book to make a persuasive case for a descriptivist approach to proper names. But when I consider the potential disorder that pure neo-descriptivism could cause in all these ‘well-established’ views about reference, even a thousand-page book defending the descriptivist-internalist-cognitivist perspective and answering all the relevant arguments still seems insufficient. And the reason is clear: most specialists are now working within a paradigm of ‘normal philosophy,’ and many do not wish to be convinced. Taking into consideration that I am not writing for readers with unshakeable theoretical commitments, in what follows I dare to offer a short summary of my own view on proper names.
1. A meta-descriptive rule for proper names
According to descriptivism, proper names are abbreviations of definite descriptions. The most explicit formulation of descriptivism for proper names – the bundle theory as presented in the work of John Searle – states that a proper name abbreviates a bundle of definite and even indefinite descriptions that constitute its whole content. The different senses we can give a proper name that we use result from our having in mind some not previously determined sub-bundle from a whole bundle of co-referential descriptions (Searle 1958). Thus, as Frege already saw, one speaker can use the name ‘Aristotle’ to mean ‘the greatest disciple of Plato and the tutor of Alexander,’ while another can use it to mean ‘the tutor of Alexander who was born in Stagira’. Both speakers can know they are referring to the same person, since they share at least one description.
In my view, the problem with this formulation of bundle theory is not that it is wrong, since in one way or another most objections to it are answerable (cf. Searle 1983, Ch. 9). The problem is that this theory is too vague, which limits its explanatory power. The descriptions belonging to the bundles are treated as if they were completely disordered. How important this is becomes apparent when we remember that the descriptions belonging to these bundles are what Wittgenstein called ‘expressions of rules’ (Regelausdrücke): description-rules that could possibly aid us to identify the bearer of a proper name. Usually there are numerous descriptions that could be associated with any proper name, many of them obviously irrelevant. Unfortunately, bundle theory has no method for deciding which description-rules belonging to a bundle have more relevance for the identification of a name’s bearer. It thus appears that the lack of such a method is the most serious flaw in traditional bundle theory.
Accordingly, my working-hypothesis is that speakers of our language implicitly appeal to some kind of general meta-descriptive rule when using a proper name. This rule should tell us the conditions under which consistency with a bundle of descriptions attached to a proper name makes this name applicable to its bearer. I intend to show that such an additional rule is possible and would greatly enhance the bundle theory of proper names.
First we must find the most relevant descriptions. My proposal is inspired by J. L. Austin’s method of quasi-lexicographical examination of ordinary language as a philosophical starting point. He recommended beginning with the Oxford Dictionary. Since dictionaries aren’t the best places to find the meanings of proper names, I suggest first looking at encyclopedia entries for proper names. We can clearly distinguish two kinds of description-rules that can help identify the bearer of a proper name. I call them auxiliary and fundamental descriptions. Fundamental descriptions are usually placed at the start of an article.
I begin with less relevant auxiliary descriptions. These can be characterized as ones only accidentally associated with proper names. Regarding the name ‘Aristotle,’ typical examples are metaphorical descriptions like Dante’s ‘the master of those who know.’ Other examples of auxiliary descriptions are ‘the greatest disciple of Plato,’ ‘the tutor of Alexander,’ ‘the founder of the Lyceum’ and ‘the man called “Aristotle.”’ These are what we may call accidental, but well-known descriptions. There are also accidental and little known descriptions associated with the name ‘Aristotle,’ such as ‘the lover of Herphyllis’ and ‘the grandson of Achaeon.’ Finally, there are contextually dependent adventitious descriptions, like ‘the philosopher mentioned by the professor in the last class.’ An adventitious description is often very transitory, as it is closely associated with an event that in most cases will soon be forgotten.
Descriptivist philosophers like Frege and Wittgenstein have often mentioned auxiliary descriptions as parts of a bundle. However, this is very misleading, since ultimately they are of negligible semantic relevance. An indication of this secondary role is shown by biographies in encyclopedias (biographies and autobiographies offer a wide range of descriptions, mostly irrelevant for identification purposes). Encyclopedias seldom begin articles with auxiliary descriptions, but instead with what I call fundamental descriptions: non-accidental descriptions that usually tell us the ‘when,’ the ‘where’ and the ‘why’ of proper-name bearers. We can define fundamental descriptions as being of the following two types:
(A) Localizing description-rule: a description that localizes an object in space and time, identifying its spatio-temporal career.
(B) Characterizing description-rule: a description that indicates what we regard as the most important aspects of the object, giving our reasons for applying the proper name to it.
Indeed, as a rule, encyclopedias first state a spatio-temporal location and then the main reasons we use a proper name; only after that do they give a more detailed exposition containing most of the auxiliary descriptions. One example is the reference to Aristotle in my short Penguin Dictionary of Philosophy, which begins:
Aristotle (384-322 BC) born in Stagira, north of Greece, Aristotle produced the major philosophical system of Antiquity…
What we first see here are in synoptic form the localizing and characterizing descriptions.
Having presented the two most fundamental kinds of description-rules, and after considering several different alternatives that I cannot go into here, I offer the following meta-descriptive rule to establish conditions of application for most if not all proper names:
MD-rule for the application of proper names:
In any possible world where a proper name ‘N’ has a bearer, this bearer must (i) belong to some most proximally relevant class C, so that it (ii) sufficiently and (iii) more than any other referent satisfies (iv) the conditions set by at least (A) its localizing description-rules and/or (B) its characterizing description-rules. (We may add to this, as helpful indicative elements, a variety of auxiliary descriptions).
I will illustrate my proposal with the name ‘Aristotle.’ The (i) most proximally relevant class C to which Aristotle belongs is that of human beings (C serves for practical reasons to narrow the scope of referents to be considered, e.g., it excludes celestial bodies or computers). The condition of type (A) for Aristotle can be summarized by the definite description ‘the person born in Stagira in 384 BC who spent the most productive part of his life in Athens, visited Lesbos and was exiled to Chalcis, where he died in 322 BC…’ The condition of type (B) for Aristotle can be summarized in the definite description ‘the philosopher who developed the relevant ideas of the Aristotelian opus…’ (That these two conditions are the most basic is supported by major encyclopedias).
Now, by applying the general meta-descriptive rule to the bundle of descriptions abbreviated by the name ‘Aristotle,’ we finally arrive at what I call its specific identification rule, the IR-Aristotle (We can also read the MD-rule simply as the form that any IR-rule for a proper name needs to establish its referential condition). Summarizing the descriptions, here is the identification rule for Aristotle:
IR-Aristotle: In any possible world where there is a bearer of the name ‘Aristotle,’ this bearer must be (i) the human being who (ii) sufficiently and (iii) more than any other satisfies (iv) the conditions of having been born in Stagira in 384 BC, spent the major part of his life in Athens and died in Chalcis in 322 BC and/or was the philosopher who developed the main ideas of the Aristotelian opus (with the possible helpful addition of auxiliary descriptions like ‘the greatest disciple of Plato,’ etc.).
We can do the same with bundles of descriptions associated with many different proper names, such as ‘Paris,’ ‘Leaning Tower of Pisa,’ ‘Amazon River,’ ‘Uranus,’ ‘Sweden,’ and, of course, also with the proper names of historically anonymous persons, though in the last case in a much more dispersive way that I cannot consider here. In what follows, I expect to show that there is nothing arbitrary about my proposal.
2. Identification rules at work
The application of the meta-descriptive rule to the name ‘Aristotle’ to obtain its identification rule enables us to give a straightforward answer to Kripke’s modal objection, according to which descriptivism is false, since no description is guaranteed to apply to any existing bearer of a proper name. As he puts it, there could be possible worlds where Aristotle lived 500 years later or where he died as a child and never wrote anything about philosophy (Kripke, 1980: 62 f.). However, these possibilities are no threat to the rule stated above, since this rule is based on an inclusive disjunction. Aristotle would still be Aristotle if he had lived 500 years later in Rome, insofar as he sufficiently satisfied the characterizing description related to his work, for instance, by writing major parts of the Aristotelian opus. He could also have died as child, as long as he sufficiently satisfied the localizing description, for instance, birth in Stagira in 384 BC as the son of the court physician Nicomachus.
Since our identification rule for Aristotle demands only sufficient satisfaction of an inclusive disjunction of the two fundamental descriptions (which purposely does not make any precise specification of degree), we can easily regard the two above considered possibilities as satisfying the identification rule.
Indeed, there are even proper names that satisfy only one description-rule of the disjunction, ‘one-foot’ (i.e. having only one of the usual two descriptions) proper names like ‘universe’ (which contains all that exists and thus can have no localizing description) and ‘Z’ (understood as the arbitrary name for the center of a circle without any relevant characterizing description). There are even proper names able to satisfy one term of a disjunction much more than the other, as in the case of a numbered planet of the solar system. Venus, for instance, must satisfy the localizing description-rule requiring that it be the second planet from the sun, orbiting between Mercury and the Earth for a sufficient period of time. Even if for some reason it changed its orbit, the essential element of its characterizing description is already built into this localizing description.
The only inconceivable alternative is that neither the localizing nor the characterizing description-rule is satisfied to a minimal degree. Such a case was fancifully suggested by John Searle in the following example:
If a classical scholar claimed to have discovered that Aristotle was no philosopher and wrote none of the works attributed to him, but was in fact an obscure Venetian fishmonger of the late Renaissance, then the ‘discovery’ would become a bad joke. (Searle 1967: 490)
Clearly, no sane person would agree with Searle’s classical scholar. Such an illiterate man could not be our Aristotle; the obvious reason is that the fishmonger does not at all satisfy either the localizing or characterizing descriptions.
Two other important elements of the MD-rule for proper names need some clarification. These are what I call the conditions of (ii) sufficiency (‘condition of being adequate or sufficient’) and (iii) predominance (‘more than any other’).
First, take the condition of sufficiency. We can imagine a possible world where Aristotle was born in 384 BC in Stagira… but died when he was seventeen, because his ship sank in a storm while crossing the Aegean to study in Athens under Plato. He may have been only an Aristotle in potentia, but we would still believe he was our Aristotle! The reason is that the identification rule is satisfied insofar as the localizing conditions are at least sufficiently satisfied (if only partly). It is irrelevant that the other term of the disjunction is not satisfied at all. The opposite case would be that of a possible world where the only Aristotle was born 500 years later than ours, lived in Rome and wrote only the Metaphysics and the Nicomachean Ethics. We would still tend to identify him as our Aristotle. We can of course imagine a possible world where the only Aristotle was born in Stagira in 384 BC, wrote Aristotle’s earlier dialogues (now lost) and the Organon, and died prematurely before writing the Metaphysics and other major works. In this case, both rules are only partially, but at least sufficiently satisfied, and we can still identify him as our Aristotle.
The second condition, predominance, reveals its purpose when we imagine that the court physician Nicomachus fathered twin sons in Stagira in 384 BC, naming both ‘Aristotle.’ The first Aristotle moved to Athens when he was seventeen to study philosophy with Plato and later wrote the entire Aristotelian opus. The second Aristotle had less luck... He became a physician like his father and accompanied Alexander on his military campaigns, but succumbed to hunger and thirst in the desert while returning from the East. Who would be our Aristotle? The first Aristotle, of course. The reason is that more than his brother he satisfies the fundamental conditions of the identification rule for Aristotle. The condition of predominance excludes the possibility that more than one referent can satisfy the identification rule. If there is more than one referent that satisfies the identification rule to the same degree, even if in different ways, our criterial tool for the application of a proper name will fail. Suppose, for example, that in a possible world a child with two identical heads was born in Stagira in 384 BC. The two heads developed into separate ‘persons’ and both were called ‘Aristotle.’ Having the same body, they inevitably lived the same life, jointly writing the entire Aristotelian opus. It would be pointless to ask which was our Aristotle, since by definition proper names apply to only one bearer, and the two satisfy the identification rule in equal measure.
Including the condition of predominance already has the advantage of explaining why it is intuitive for us that a Twin-Earth Aristotle (who was qualitatively identical with our Aristotle and lived in a similar world) is not the ‘true’ Aristotle. This approach works better than Searle’s attempted explanation (1983: 254-5), because our earth’s Aristotle better satisfies the condition of predominance. Both Aristotles satisfy the characterizing description rule (both wrote the Corpus Aristotelicum), and because the spatial context is similar, the Twin-Earth Aristotle also appears to satisfy the localizing rule. But beyond this, only the Earth-Aristotle really satisfies the localizing description-rule, since he lived in the Greece of our Earth, and not in the far distant Twin-Earth Greece. Because both Earths belong to one and the same space, the localizing description-rule refers to a spatial location on our Earth and not to its counterpart on the distant Twin-Earth, notwithstanding the confusingly similar local surroundings. (Even if the spaces were incommensurably different, the conclusion would remain the same, since our Aristotle belongs to the first and not to the second space).
The introduction of the so understood identification rule allows us to solve the famous paradox of Theseus’s ship. Suppose Theseus had a ship ‘Calibdus’ that over time showed signs of wear. He gradually replaced its planks with new ones, until in the end there was not one original plank left. All the worn-out planks had been stored, and someone decided to repair them and build a ‘new’ ship, identical with the original one. Which ship is now the Calibdus? (If you think it must be the first one, you need only speed up the substitution of the planks: if the whole substitution requires just one month, you would tend to say that the second ship is the Calibdus, and if it takes just one day, you will be sure of this.)
This seems paradoxical, and we can imagine a contradiction in the application of the two fundamental descriptions. The first ship better satisfies the localizing description concerning its date of launching and spatio-temporal career; it also satisfies enough of the characterizing description concerning its structural and functional properties, though not its material constitution. The second ship satisfies the characterizing description better, since besides its structural and functional properties it has all the original parts. Both satisfy conditions of sufficiency for Calibdus, but since we are invited to imagine a situation in which neither of them satisfies the condition of predominance, the identification rule isn’t applicable any longer, and thus Calibdus no longer exists in the sense of the word meant here.
One could finally ask if auxiliary descriptions play some role regarding predominance. The answer seems to me double. The answer is ‘no’ in cases where auxiliary descriptions are irrelevant, like ‘the man called by Dante the master of those who know’ or ‘the grandson of Achaeon.’ But in the case of relevant auxiliary descriptions like ‘the greatest disciple of Plato’ they seem to matter. And one reason is that the border between fundamental and auxiliary descriptions is vague, particularly for proper names of historically unimportant persons.
Finally, the insignificance of auxiliary descriptions comes to the fore when we consider the case of someone who satisfies them but does not satisfy the fundamental conditions. Consider, for instance, the millionaire Greek shipping magnate Aristotle Onassis (1906-1975). He could not be our Aristotle, because he satisfies none of the fundamental descriptions. But suppose he satisfied some auxiliary descriptions of our bundle, say ‘the man called “Aristotle,”’ ‘the tutor of Alexander,’ ‘the master of those who know’ and ‘the lover of Herphyllis.’ This changes absolutely nothing in our judgment! Although his name was ‘Aristotle,’ he could not be the greatest philosopher of ancient Greece. He could not be, even though he did in fact educate his son Alexander, because this son could not be the greatest conqueror of Antiquity. Even if someone called him ‘the master of those who know,’ it does not matter, as that person would surely not be Dante. And he could not, even if he had a mistress named Herphyllis, have a relationship with the famous concubine from ancient Stagira. It does not matter how many auxiliary conditions related to our true Aristotle this proper name satisfies, they will never suffice to identify him. We regard them as irrelevant and surprising coincidences, showing that auxiliary descriptions can only be helpful – as we will see – insofar as the fundamental descriptions are already applicable.
3. Objection of vagueness
At this point one could object that identification rules derived from the MD-rule (or instantiating it) are too vague. Not only do they not establish precisely how much of an inclusive disjunction must be satisfied in order to be sufficient, they do not establish precisely how much more a possible competitor must satisfy the disjunction in order to disqualify another with certainty.
To answer this question we need to begin by remembering that vagueness does not mean (as shown, e.g., by the sorites paradox) that boundaries disappear. After all, it is quite easy to imagine a possible world where we cannot be certain whether or not our Aristotle ever existed. This would be the case, for example, in a possible world where Aristotelian philosophy never developed, but there was a court physician in Stagira named Nicomachus who in 384 BC fathered a child with a congenital birth defect that died soon after birth. The parents even named this baby ‘Aristotle’... Would the child be our Aristotle? We cannot tell.
Having this in mind, the objection is easy to answer. Our natural language is vague; for our semantic rules to be truly applicable in other empirically possible worlds, they must leave room for vagueness. This is precisely what our identification rules do. Thus, far from being a problem, their vagueness is advantageous. It is evidence for their correctness, since all their vagueness does is to mirror the semantic ambiguity of our own natural language. In our example, this is shown by the congenitally defective Aristotle, for whom the condition of sufficiency cannot be met because of unavoidable vagueness.
Saul Kripke correctly classified proper names as rigid designators, defining a rigid designator as a term that designates the same object in every possible world where this object exists or could exist, designating no object in a world where this object does not exist (1971: 145-6). So understood I see this as an intuitively useful idea for the selection of the most adequate theory of proper names. From this idea we can derive the following well-known rigidity test: If you wish to know if x is a rigid designator, you must ask if x could have existed without being x (Hughes 2004: 20).
Now, since in some possible worlds there are cases where we cannot know whether the bearer of a proper name exists, we must redefine the rigid designator as the term that applies in every possible world where its reference definitely (unambiguously) exists. So understood, the identification rule proposed above again makes the proper name a rigid designator – a point to which we will return below.
The proposed meta-descriptivist theory of proper names can help us gain a better understanding of proper names’ meanings. Leaving aside for now what individual speakers can mean with a name – which is variable and often very limited – we can say that the core meaning (sense) of a proper name is given by its fundamental description-rules. The reason for this is that together with the entire identification rule they are able to singularize the name’s meaning by identifying its sole bearer. These fundamental descriptions must be conventions that are sufficiently known, at least by what we may call privileged speakers (in many cases ‘specialists,’ they may even be only partially shared by them), understood as those who (alone or jointly) are truly able to apply them in making an identification. So, if you know that Aristotle was ‘the philosopher who wrote the Metaphysics’ and that he was ‘a person born in Stagira in 384 BC who lived most of his life in Athens,’ you already have some decisive informational meaning, though not if you only know that he was ‘the tutor of Alexander.’
Now, what about auxiliary descriptions? They are still able to give an aura of meaning to a name, which sometimes becomes apparent, as in the case of ‘Plato’s greatest disciple.’ Nonetheless, someone who only knows a supplementary description usually associated with a proper name, like ‘the teacher of Alexander’ in association with ‘Aristotle,’ whom he saw portrayed by an actor in a movie about the life of Alexander, does not really know anything relevant about the meaning of the name Aristotle. Yet, even if he only knows a little meaningful information about him, he can use this knowledge to insert the name correctly in some sort of vague discourse, making what I call a parasitical kind of reference. Auxiliary descriptions have an auxiliary role of pointing, guiding a speaker within a linguistic community, directly or indirectly linking him with a chain to other speakers who know the fundamental descriptions and eventually enabling him to identify the bearer (this is sometimes called a process of ‘reference borrowing’).
Finally, here we should not confuse cognitive with emotive meaning. The bundle of descriptions associated with a proper name, particularly regarding fundamental descriptions, gives its informative or cognitive content – what Frege called its sense (Sinn). This has a conventional ground that is in some way implicitly or explicitly established as something shareable between speakers. However, there are also things like images, memory-images, feelings, smells … that can be strongly associated with a proper name (e.g., ‘the Pietà,’ ‘Gandhi,’ ‘Stalin,’ ‘Auschwitz’…), but cannot be easily captured by descriptions. We could say they belong to an imagistic-emotive dimension of meaning, which would be based on the often-shared regularities of our psychological reactions instead of our usually explicitly established conventions. I believe that the widely disseminated idea that not all our cognitive meanings can be linguistically expressed in the form of descriptions arises from a failure to distinguish imagistic-emotive senses from conventional meanings. Because it is descriptively expressible and conventionally grounded, cognitive meaning has a shared basis that allows hearers to decode it.
5. Ignorance and error
Possessing this general explanation of the meaning of proper names, we are prepared to give an answer to Kripke’s counter-examples of ignorance and error. They concern people who associate an indefinite description with a proper name, such as ‘a physicist or something like that’ with the name ‘Feynman.’ They also concern people who associate erroneous descriptions with a proper name, such as associating ‘the developer of the atom bomb’ with the name ‘Einstein’ or ‘the originator of Peano’s axioms’ with the name ‘Peano’ (actually these axioms were originally discovered by Dedekind and were only refined by Peano) (Kripke 1980: 81-89).
My answer is that the speaker is able to endow proper names like these with a merely parasitical or borrowed referential role. To do this, it suffices to know a very marginal and insufficient description, as long as we are sure the linguistic community includes privileged members with the necessary knowledge of the fundamental description-rules to really apply the identification rule. This means that a speaker who knows such descriptions is already able to insert a word into discourse in a way he expects can associate the name with its bearer somewhere in the communication network. Important for the success of this parasitical form of reference is that the description known by the speaker enables him to insert the proper name in an understandable way into sufficiently vague discursive contexts. This is the case of the Kripkean counter-examples presented above. One can correctly insert names in a sufficiently vague discourse by associating them with even just one indefinite or erroneous description, insofar as at least the following two conditions for parasitical reference are satisfied:
(a) The description known by the speaker is convergent. That is, a description that at least correctly classifies the name’s owner (class C of the identification rule).
(b) The speaker implicitly knows the MD-rule for proper names. This means he must at least be aware that he does not know more than an irrelevant part of the meaning, which will make him careful enough when inserting the name in discourse (he knows how little he knows).
To give a simple example: Not being a physicist, I know very little about the cognitive meaning of the abstract name ‘string theory.’ But at least I know how little I know. I could even fool my students by giving some vague information about super-strings as incredibly small vibrating filaments of energy that produce all the matter and energy in the universe by vibrations of different frequencies… The scientific context allows this, although in fact I am far from understanding any mathematical ideas constitutive of the theory. This is why in a more demanding context, for instance a discussion among theoretical physicists, I would not try to participate. Furthermore, without these privileged speakers and their adequate knowledge of meaning, my insertion of the word ‘string theory’ into discourse would be vacuous, for the parasitical reference would have nothing to attach itself to. If all specialized knowledge of string theory should disappear due to a cosmic catastrophe, the meaning of this name would also be lost, even if someone could still remember how to spell it.
Consider now Kripke’s counter-examples. A person can insert the name ‘Feynman’ in sufficiently vague discourses. Her use must be convergent, she must correctly classify Feynman as ‘a physicist or something like that’ and therefore as a human being, and she must be implicitly aware of the MD-rule. A person can also use the names ‘Einstein’ and ‘Peano’ correctly in vague discursive contexts, possibly expecting to obtain more information or even correction from better informed speakers. She must simply satisfy conditions (a) and (b), correctly classifying Einstein as a theoretical physicist, Peano as a mathematician and both as human beings.
On the other hand, when proper names are associated with divergent descriptions, thus incorrectly classified, the referential thread is apt to be lost. Thus, if speakers associate the name ‘Feynman’ with the divergent description ‘a brand of perfume,’ the name ‘Einstein’ with the divergent description ‘a precious stone,’ and the name ‘Peano’ with the divergent description ‘a musical instrument,’ they will probably not be able to insert these names correctly in any discursive context, no matter how vague it may be. We will not say that in using the name they are able to refer to its bearer, even in a borrowed or parasitical way.
Curiously enough, the same conditions also apply to general terms. If a fisherman means by a whale a large marine fish, this is incorrect, as whales are mammals, but at least it is convergent, since he classifies the whale correctly as a sea animal, which already enables him to insert the word in colloquial discourse. However, if a child believes that ‘whale’ is the name of a mountain in the Appalachians, his usage is not only incorrect but also divergent, making him unable to adequately insert the word into discourse.
Finally, I can use what we have learned to refute a counter-example to descriptivism suggested by Keith Donnellan (1972: 374). He describes a case in which a close friend, Tom, visits a couple and asks to see their child, who is asleep in his bed upstairs. The parents agree, awaken the child and introduce their friend, ‘This is our friend Tom.’ Tom greets the child with ‘Hello,’ and the child, after hearing this, immediately falls asleep again. Asked about Tom the next morning, the child replies, ‘Tom is a nice person,’ without associating any definite description with Tom. Even though he would most likely be unable to recognize Tom on other occasions, according to Donnellan he has still succeeded in referring to Tom!
Our answer can vary depending on the details of the story. If the child has no memory of being awakened, of having seen or heard anyone, then he is only trying to satisfy his parents. In this case, of course, he is not actually referring to anyone. However, suppose the child still has some vague memory of seeing a strange person the previous night. If he saw Tom on the street, he would not recognize him. Nevertheless, in this case he is already using the proper name in a convergent way, since he associates the name ‘Tom’ with the description ‘a friendly person whom I saw last night.’ In this particular discursive context, hearers who know the identification rule for Tom will be able to give the utterance its full meaning. The parents are privileged speakers here. They know Tom’s appearance, what he does for a living, where he lives, where he comes from and many other details of his life. Indeed, without this additional knowledge, the child’s comment would be empty, not really referring to a particular Tom in any satisfactory sense. The child’s vague reference to Tom must be supplemented by his parents, who know the causal circumstances and are able to refer to Tom in the full sense of the word by means of his name’s identification rule.
From a descriptivist perspective, our neo-descriptivist theory explains why proper names are rigid designators. For us proper names are rigid designators because their identification rules apply in any possible world where the proper name’s bearer exists. This is the real reason they satisfy Kripke’s condition, according to which ‘a’ is a rigid designator if and only if it is false that some ‘a’ might not have been ‘a.’
Moreover, I have a descriptivist explanation for this. What this really means is that a name’s bearer cannot exist without satisfying the identification rule for this bearer, as this rule simply defines what this bearer can be in any possible world. The identification rule establishes all the possible combinations of descriptions of particularized properties that a referent must have in order to be the sole bearer of its proper name.
To clarify this point we can express a proper name’s identification rule in the form of a definitional identity sentence able to identify the bearer of the proper name by means of a complex associated definite description. As an example, we need only formulate the rule identifying the proper name ‘Aristotle’ with a complex definite description as follows:
Aristotle (Df) = the name that in any possible world where it has a bearer applies to a human being who sufficiently and more than any other satisfies the condition that he was born in Stagira in 384 BC… died in Chalcis in 322 BC and/or was the author of the key ideas in the Aristotelian opus.
This formulation of the identification rule for Aristotle is an analytically necessary a priori statement. It contains the complex definite description ‘the name that in any possible world where… ,’ which besides defining what the name means is also a rigid designator. It passes the proposed rigidity test: the name ‘Aristotle’ is rigid because one cannot imagine a possible world where Aristotle exists but is not Aristotle because he doesn’t satisfy the above definition.
We can see that, unlike the old descriptivism, the meta-descriptivist view does not risk destroying the rigidity of proper names. Quite to the contrary, it enables us to show their rigidity descriptively, since it explains the conditions under which any possible world may be home to the bearer of a name, to whom the name necessarily applies. The reference occurs by means of particularized properties or tropes (see Appendix to Chapter III), insofar as they satisfy criterial configurations that can be generated by the rule and are seen as sufficient for its application. However, the particularized properties that satisfy the respective criterial configurations are not always the same. They can change in multiple and varied ways, constituting no permanent individualizing essence. Thus, in one possible world we can identify Aristotle as a person born in 384 in Stagira the son of Nicomachus, and in another possible world we could identify Aristotle as the person who wrote the Metaphysics and the Organon… This flexibility of the identification rule frees us from having to include essential properties of the referenced object that are seen as necessary and sufficient conditions for the name’s application.
One objection was that since conventional rules are changeable this endangers rigidity. Consider the following supposed counter-example. It is well known that early in his life – a period I call ∆t1 – David Hume was known as a major historian but not as a philosopher, and thus our characterizing rule for him could be ‘the author of The History of England.’ I call this corresponding early identification rule IR-Hume1. Now, at a later time and up and until the present – a period I call ∆t2 – Hume became much better known as ‘the author of the Treatise’ rather than ‘the author of The History of England.’ I call this present identification rule containing more information IR-Hume2. Now, suppose that in a future time ∆t3 the information that Hume was the author of The History of England is completely lost and the only remaining characterizing description is ‘the author of the Treatise.’ I call this identification rule IR-Hume3. Comparing IR-Hume1 with IR-Hume3, we see a case in which the available identification rule completely changes. Now imagine there is a possible world where there is a single Hume who only wrote The History of England and another possible world where there is a single Hume who only wrote the Treatise. In this case, we would apply IR-Hume1 to the historian of the first world and IR-Hume3 to the philosopher of the second world, perhaps identifying different persons in different possible worlds. Now, imagine a bizarre situation. Suppose that in a possible world wt, very similar to ours, there were identical twins, the Humes, who had insufficiently distinguishable localizing descriptions, but one was only the historian, while the other only wrote the Treatise. In ∆t1 using the rule IR-Hume1, we would identify the first person as our Hume; in ∆t3, using IR-Hume2, we would identify his twin. This shows that changes in the identification rule can lead us to identify different objects in the same possible world, destroying the rigidity of the proper name. The only way to preserve rigidity is to agree that what a proper name means must be what is meant according to an identification rule set by privileged speakers of a language community at some ∆t. This is, for instance, the case of our own IR-Hume2, which due to the condition of predominance identifies the writer of the Treatise in wt as our Hume and not his twin, who only wrote The History of England.
It is worth noting that IR-Hume2 exemplifies an increase in the number of details of meaning, making the rule more precise. With IR-Hume2 we have more elements with which to identify the same object, and we would still identify the same object in all possible worlds. Now, however, we would also identify it in other worlds where we couldn’t definitely identify the object by using IR-Hume1 alone. The example of a transition from IR-Hume1 to IR-Hume2 is important because it is very common: normally our information about a proper name’s owner increases with time (think of very detailed biographies). That is, over time we usually add descriptions to a core, making the boundaries of its application sharp enough to decide on doubtful cases that earlier lay within the blurred borders of application. However, this is not sufficient to destroy rigidity, since because of this we do not choose different objects in possible worlds where the object exists, but only improve our acuity. Hence, I conclude that by accepting an increase of identifying conventions we are not forced to abandon rigidity. Furthermore, I suspect that the amplification and change of descriptions associated with a proper name would cause much more trouble for Kripke’s view, albeit hidden by his coarse-grained analysis. For by what means could he identify the right Hume in wt, except by implicitly considering him first as ‘the writer of the Treatise,’ namely, the person who satisfies our IR-Hume2?
An example illustrating a related point concerns the name ‘Madagascar,’ suggested by Gareth Evans against Kripke’s causal-historical view of proper names (Evans 1973). ‘Madagascar’ was initially used as the name of eastern regions of Africa. During his world travels, Marco Polo visited a large island off the coast of eastern Africa and mistakenly began to use the name Madagascar for it. Today, because of Marco Polo’s mistake, we also call this island Madagascar. However, if Kripke’s theory were correct, according to which the reference of a name is fixed by a causal chain beginning with its first baptism, we would still use the name Madagascar for the eastern part of Africa. Kripke tried to solve this problem by suggesting there is a new social intention to refer to the island that overrides the former intention. This dangerously approaches a recognition of the necessity of new descriptions (disguised as intentions) to identify the island (Kripke 1980: 163.
From our perspective, we can easily solve the problem. We could admit that this is a case of homonymy, since there is a forgotten Madagascar-1 of eastern Africa, with its proper fundamental descriptions, and the well-known Madagascar-2, the island, with very different proper fundamental descriptions. Here we have a new identification rule created for a new reference object.
7. Names versus descriptions
Perhaps the greatest advantage of my proposal is that it gives a much better explanation of the contrast between the rigidity of proper names and the accidentalness/flaccidity of definite descriptions. According to Kripke, unlike proper names, definite descriptions can have different bearers in different possible worlds. So, while the name Benjamin Franklin always refers to the same person in any possible world where this person exists, the description ‘the inventor of bifocals,’ which refers to him in our world, could refer to a different person or even to no person in some other possible world. For Kripke this is the case because proper names, being rigid, are after their bearers’ baptism necessarily linked with them in a mysterious way that calls for explanation. Definite descriptions belong to a different epistemic category. We could say that they refer by means of what John Stuart Mill would call their connotation, defined by him as the implications of attributes belonging to the referenced object (Mill 2002: 19, 21).
Nonetheless, in making this sharp distinction Kripke overlooked the most relevant point, that definite descriptions are only accidental in relation to proper names. The consequences can be shown first intuitively and then using Wittgenstein’s distinction between symptoms and criteria.
Intuitively, the reason why most definite descriptions are accidental designators (such as ‘the inventor of bifocals’) is that when we apply them we associate them semantically, in a contingent way, with the identification rule of some proper name (such as ‘Benjamin Franklin’). Indeed, this association isn’t normally established as necessary by identification rules (and our MD-rule). Consequently, we can easily imagine possible worlds where there is a mismatch between the object possibly referred to by a proper name and the object possibly referred to by the definite descriptions usually attached to it, particularly when these descriptions are merely auxiliary ones (for example, in a world where Samuel Adams invented bifocals and Benjamin Franklin never lived).
We can elaborate our explanation of the distinction between the rigidity of proper names and the accidental character of descriptions with the help of Wittgenstein’s distinction between symptoms and criteria (see Chap. II, sec. 7 and Chap. III, sec. 10 of this book). According to this distinction, once accepted as given, a criterion warrants the application of a word, while a symptom, once accepted as given, makes this application only more or less probable. In their association with proper names, definite descriptions typically give us only symptoms for their application, particularly when they are auxiliary, though sometimes even when they are fundamental. This explains why these descriptions alone are not applicable in all possible worlds where the bearer of a proper name exists. By contrast, the complex definite description expressing the identification rule of a proper name is able to generate multiple criteria to identify the referent (e.g., Aristotle) in different possible worlds. These criteria can be met by particularized properties (tropes) like those satisfying the examples given above. One can say that in different possible worlds the bearer of a proper name can satisfy the same identification rule in different ways, by means of many different possible configurations of particularized properties.
An easy way to prove my reasoning is correct is by explaining a phenomenon that Kripke’s causal-historical view cannot explain. We simply have to find definite descriptions that are not semantically associated with any proper name. In this case, we expect them to behave as rigid designators, applying to only one object in any possible world where this object exists. I call them autonomous definite descriptions. The following four descriptions are examples:
1. the 52nd Regiment of Fot,
2. the last living Neanderthal man,
3. the 1914 assassination of Austrian Archduke Ferdinand in Sarajevo,
4. the easternmost point of South America.
These descriptions respectively name a military organization, a human being, an event, and a place. It is important is to note that they are all easily recognized as rigid designators. (1) We can imagine a similar possible world where the 52nd Regiment of Fot had a different organization and time of existence, for example, where it did not serve in the Napoleonic wars. (2) We can imagine a possible world where the last Neanderthal man outlived all members of homo sapiens. (3) We can imagine a world where the Archduke was assassinated at a different time by someone other than Gavrilo Princip. (4) And we can imagine a possible world where the easternmost point of South America is not in Brazil but in Tierra del Fuego, which stretches very far eastward toward Africa (assuming that we consider ‘the same point’ regardless of properties and latitude). Even so, if applicable these descriptions will always be applicable to the same bearer in every possible world where this bearer exists, whether it is an organization, a human being, an event or a geographic location. These definite descriptions are rigid designators, simply because the identification rules made at least partially explicit by them (with their localizing and/or characterizing description-rules) are always able to pick out the same referent, without the danger of mismatching with referents picked out by the identification rules of associated proper names. The Kripkean view has no explanation, except the ad hoc claim that autonomous descriptions are nothing but disguised proper names.
8. Autonomous definite descriptions
Finally, it is worth noting that the same MD-rule we apply to bundles of proper names could be applied in the case of autonomous definite descriptions, insofar as they are seen as rigid designators and singular terms independent of any proper name. The difference is not just that part of the rule usually belongs to its implicit content (as a ‘connotation’), but also that these rules are often less complex. I can give as an example the identification rule for the ‘the 52nd Regiment of Fot.’ It has the following (summarized) localizing rule:
The 52nd Regiment of Fot existed from 1757 to 1881; it saw active service particularly during the American War of Independence, the Anglo-Mysore wars in India and the Napoleonic Wars.
The identification rule for the 52nd Regiment of Fot has the following (summarized) characterizing rule:
It was a highly regarded regiment whose troops were drawn chiefly from Oxfordshire, consisting of one or two battalions of light infantry, each comprising approximately 1,000 men.
Of course, the inclusive disjunction of these descriptions needs to be only sufficiently and predominantly satisfied in any possible world where ‘the 52nd Regiment of Fot’ exists. Auxiliary descriptions are also present, for instance ‘a regiment never surpassed in war since men first took up arms,’ but are of little relevance. The same is the case with other autonomous definite descriptions.
On the other hand, definite descriptions like ‘the inventor of bifocals’ or ‘the tutor of Alexander,’ employed in close association with proper names, are viewed as merely auxiliary descriptions because of their connotations. They are poor complements to the identification rule of their associated proper names.
9. Some classical counter-examples
It is a highly useful notion that the meta-descriptivist rule must be applied to the bundle of descriptions associated with any given proper name in order to construct the proper identification rule. Using this tool, we can go beyond the causal-historical sketch, and also rehabilitate descriptivism, giving more satisfactory answers to objections. We see this in its capacity to answer Kripke’s modal objections, according to which descriptivism is wrong because any description or group of descriptions associated with a name can fail to refer to the name’s bearer. But, if considered as rigid designators, proper names never fail to refer. To clarify this position, I will discuss four modal counter-examples.
(i) The first is Kripke’s memorable Gödel counter-example (1980: 83-84). Suppose Mary knows nothing about Kurt Gödel, except the description ‘the discoverer of the incompleteness theorem.’ Then suppose that in nineteen-thirties Vienna an unknown Viennese logician named Schmidt wrote the first paper to describe the incompleteness theorem but died before he could publish this major discovery. Later, a certain Gödel found his manuscript and published it under his own name. According to Kripke, if the descriptivist theory were correct, Mary should conclude that ‘Gödel’ means the same thing as ‘Schmidt.’ But it is obvious that the name ‘Gödel’ still refers to the real Gödel! And according to Kripke, the reason is that the reference is fixed by the baptism of the infant Gödel. This is followed by a causal-historical chain in which each hearer repeats the name with the intention to refer to the same person referred to by the speaker from whom he heard it, continuing up to Mary’s utterance…
However, this objection poses a threat only to Kripke’s own caricatured formulation of descriptivism. Our identification rule for the name ‘Gödel’ goes much farther. First, the characterizing description-rule for the name ‘Kurt Gödel’ can be summarized as:
a great logician who made major contributions to logic, particularly the incompleteness theorem.
This is already more than what Mary knows, since this description also points to Gödel’s other contributions to logic. Moreover, Kripke does not even consider the localizing description, which can be summarized as:
the person born in Brünn in 1906 who studied in Vienna, emigrated to the USA in 1940 via the trans-Siberian railway, and worked at Princeton University until his death in 1978.
As a competent speaker of the English language, Mary must unconsciously know the MD-rule. She must be tacitly aware that to conclude that Gödel was Schmidt, she would have to do much more than just attribute the discovery of the incompleteness theorem to Schmidt. Consequently, she wisely refrains from concluding that Gödel is Schmidt.
Moreover, for a privileged speaker Gödel cannot be Schmidt, because even if Schmidt satisfies part of Gödel’s characterizing description, Gödel continues to satisfy the whole localizing description and at least part of the characterizing description, satisfying in this way the condition of predominance. Nevertheless, we can see that something in the meaning of the name ‘Gödel’ is attached to the name ‘Schmidt,’ which would be clear if someone heard a mathematician who, scandalized by this idea, angrily protested: ‘No! The true Gödel was Schmidt!’
Moreover, under some circumstances Gödel could really be Schmidt. Suppose that Schmidt killed Gödel when he was a teenager and assumed his identity. Then Schmidt studied mathematics in Vienna, conceived and published the incompleteness theorem, married a woman named Adele, moved to the USA in 1940 and worked at Princeton University until his death in 1978. In this case, we would all agree that Gödel was in fact Schmidt, the unscrupulous murderer. And the famous photo of Gödel with Einstein would actually be a photo of Schmidt with Einstein. But why should we say this? Our answer is clear: because Schmidt now satisfies the condition of predominance. He sufficiently satisfies the localizing and characterizing description-rules for the name ‘Gödel’ much more than the unfortunate teenager whose name he stole. According to Kripke’s theory, however, we should continue using the name ‘Gödel’ in referring to the person who was baptized with this name at the beginning of the causal-historical chain. That person is the murdered Gödel, and not Schmidt, independently of the fact that the first person does not sufficiently satisfy the main descriptions. But this runs counter to our intuitions.
(ii) Now consider the case of semi-fictional names like Robin Hood. From my perspective, if a name really is semi-fictional, it must be associated with some descriptive content effectively applicable to a real owner, along with merely imaginary descriptive content added later, even if we are unable to definitely distinguish one type of content from the other. – If they lack any real content that we consider applicable, they should be called ‘purely fictional names.’ Thus, with regard to these names our situation is one of uncertainty and insufficient knowledge. In the case of Robin Hood, the vague descriptions ‘a person who probably lived in England in the 13th century’ and ‘a legendary righter of wrongs’ respectively suggest some contents of almost completely unknown localizing and characterizing descriptions.
According to Kripke the story is different. It does not matter whether a semi-fictional name has any true descriptive content. Important is only that the name meets its own requirement of coming at the end of an external causal-historical chain linking it with the baptism of its reference. Hence, independently of any bundle of descriptions known or unknown to us, if this condition is met, the reference of a semi-fictional name is warranted.
I think my answer is intuitively more balanced and complete. As descriptivists, we should admit that what we think is a semi-fictional name can in fact be purely fictional. We suspect that the name has a reference, since there are hints that it could refer to a real historical person, and a solution to our puzzle could be found. Suppose historians discover documents about a man named Robart Hude, an early 13th century outlaw who championed the weak against the powerful and hid with a band of followers in Sherwood Forest near Nottingham, suggesting that his life story may have given rise to the legend of Robin Hood. With this in mind, we have enough information to apply both the correct localizing description – early 13th century, lived near Nottingham – and the correct characterizing description – an outlaw who stole from the rich and gave to the poor, originating the legend of Robin Hood. This would give us an improved descriptivist confirmation of the origin of Robin Hood as a ‘real’ semi-fictional character, while the causal-historical ‘explanation’ could not change anything.
We can also find cases suggesting inadequacies of Kripkian explanations. A scholar might discover there really was an historical model for the first medieval writer who wrote about the legend of Robin Hood, but that none of the traditional descriptions apply to it. Perhaps there was a faithful hunting dog called Robin who tagged along when the author went hunting in Sherwood Forest. Touched because the dog was always ready to help his master, the author simply made up the story of Robin Hood. Perhaps a Kripkean philosopher would conclude that Robin Hood was the dog’s name. A historical chain began when the writer baptized the dog, and all subsequent hearers shared an intention to refer to the subject first referred to as Robin Hood. But this probably strikes most readers as more than a bit strange.
On the other hand, our MD-Rule allows us to explain the case more clearly and persuasively. This rule would indicate that Robin Hood was the name of a purely fictional character and has nothing to do with any dog, since according to the identification rule, the bearer of the name ‘Robin Hood’ should at least belong to class C of human beings.
(iii) Another counter-example is that of Hesperus (1980: 57-58). Suppose, says Kripke, that someone once fixed the reference of Hesperus by using the following statement (i) ‘I shall use “Hesperus” as the name for the heavenly body appearing in that particular position in the sky.’ This does not mean that to be in a certain position in the sky is a necessary property of Hesperus. If long ago a comet had collided with the planet Hesperus, it might no longer have been in its original position when first discovered. Nevertheless, Hesperus would still be Hesperus, since the name is a rigid designator. Kripke’s bundle theory is unable to explain this case.
Our answer comes from an analysis of the identification rule for Hesperus (the Evening Star). Although one can naively define it as the most brilliant celestial body in the evening sky, it is more insightful to define Hesperus as an appearance of Venus, and I think that Kripke had the last definition of its meaning (implicitly) in mind, which he confuses with the arbitrary tagging (i). Now, the real identification rule for Hesperus can be summarized as follows:
IR-Hesperus = the celestial body that satisfies sufficiently and more than any other the condition of being the second planet from the sun in the solar system, orbiting between Mercury and Earth; it also appears as the brightest celestial body in the evening sky.
This is a case of a one-foot identification rule for which only the localizing description really counts. First, because the essential feature of the characterizing rule, that it is a planet and a very bright celestial body appearing in the evening sky (as the name ‘Hesperus’ indicates), is already built into the localizing rule. Second, because other properties of Venus, insofar as they do not prevent the application of the identification rule, are irrelevant. Moreover, this rule is a rigid designator: there can be no possible world where this Hesperus is not the second planet of the solar system, in some way appearing in the evening sky. The object can satisfy the rule only partially, although sufficiently and more than any other.
Now, suppose the planet Hesperus collided with a comet sometime after it was identified, and this collision changed its position in the sky or even destroyed it. IR-Hesperus would remain valid, even if not completely satisfied. Kripke’s example is more sophisticated. He invites us to imagine a possible world wh (which he conceives as our own world in counterfactual circumstances) with no evening star and no second brightest planet defined as such in the relevant historical time period. However, (we may suppose) astronomers on wh have discovered that there was once a second planet orbiting between Mercury and Earth that was struck by a comet and no longer exists – a planet that appeared in the evening sky, which we may call ‘Hesperus.’ I think, considering that, at least partially and more than any other, this planet satisfies the IR-Hesperus, it still could be called ‘Hesperus,’ which shows that Hesperus satisfies our identifying rule as a rigid designator even in possible world wh.
(iv) Another of Kripke’s objections – circularity in names like Peano and Einstein – is easy to answer. Limiting myself to the first, Kripke’s view is that we define the name ‘Einstein’ descriptively as ‘the originator of relativity theory,’ but we explain relativity theory as a theory created by Einstein, which leads to circularity. The answer is not just that it isn’t necessarily so (we can explain the theory without mentioning its originator’s name), but that the use of its originator’s name in its definiens is perfectly adequate; for it is natural to re-utilize a defined definiendum in the search for a more complete definiens. This reutilization is not circular; it can be part of an ‘ascending bascular movement’ in which already found information is used to obtain more information (any Google search should convince you of this).
(v) Now I want to briefly analyze some counter-examples proposed by Keith Donnellan (1970, sec. x). One instructive counter-example is the following: Suppose, he writes, someone discovers Thales was actually no philosopher, but instead a wise well-digger living in Miletus, who, exhausted by his hard work, once said ‘I wish all were water, so I wouldn’t have to dig these damned wells.’ Now, suppose this sentence came down to Herodotus, Aristotle and others in an altered form as the view of the first Greek philosopher Thales that water is the principle of all things. Donnellan adds to this story the assumption that there really was a hermit who thought all is water. However, he lived in a period so remote that neither he nor his doctrines have any historical connection with us. We wouldn’t say that the hermit was Thales, even if the hermit really satisfied the description. The reason, according to Donnellan, is clear: Thales and not the hermit was at the start of the causal-historical chain.
The answer offered by our neo-descriptivist view is that in some cases the description of a causal history is so important it must be included in the characterizing description-rule. This is precisely the case with Thales, because what we find important about him is that he came at the start of Western philosophy. Without knowing this historical context, the statement ‘Water is the principle of all things’ would seem absurd. Thus, we could summarize the real characterizing definite description belonging to the identification rule for Thales as:
the person who originated the doxography found in Aristotle and others, which describes him as having been the first Greek philosopher, who said that water is the principle of all things, that everything is alive, etc.
As for the localizing description, we at least know that Thales was:
a Milesian who lived from 624 to 547-8 BC and probably once visited Egypt.
In view of this, if we return to Donnelan’s example we must conclude that according to our version of descriptivism the hermit could not have been Thales. And the reason is that Thales the well-digger better fulfills both fundamental conditions, in this way satisfying the condition of predominance. Let us compare the two cases. The hermit does not satisfy any part of the localizing description; all he satisfies is an incomplete part of the characterizing description. On the other hand, Thales the well-digger completely satisfies the localizing description, because he lived in Miletus from 624 to 547-8 BC. And regarding the characterizing description, even if Thales were not a philosopher and never said the principle of all things is water, he remains the person wrongly described in the doxography as the first Greek philosopher who said all is water. Hence, despite everything, our Thales satisfies the fundamental descriptions much better than the hermit, thereby qualifying as the name’s logical bearer.
Aside from that, one should not forget that depending on details we could add to or subtract from this example, our intuitions would change, leading us to think our Thales didn’t really exist or even that the hermit was the true Thales.
(vi) Another of Donnellan’s counter-examples is a student who talked with a person at a party he believed was the famous philosopher J. L. Aston-Martin, author of ‘Other Bodies.’ Although the person’s name actually was Aston-Martin, he only pretended to be the philosopher. Donnellan notes that the sentence (a) ‘Last night I spoke with Aston-Martin’ is false, because it associates the name ‘Aston Martin’ with the description:
D1: the philosopher who wrote ‘Other Bodies.’
In contrast, these sentences are true (b) ‘At the end of the party Robinson stumbled and fell on the floor at the feet of Aston-Martin’ and (c) ‘I was almost the last person to leave; only Aston-Martin and Robinson were still there.’ This is because they are associated with description D2: ‘the man named Aston-Martin whom I met at the party.’ The objection is that descriptivist theory does not explain this change: in (a), (b) and (c), the name ‘Aston-Martin’ should be associated with the same bundle of descriptions that includes ‘the author of “Other Bodies.”’
I have no problem with this example. One can always attach a false description to a proper name, confusing it with a description associated with a person of the same name. The student had the wrong characterizing description, ‘the author of “Other Bodies”’ and two correct adventitious auxiliary descriptions. He really did not know the identification rule for Aston-Martin. But since he had (a) convergent descriptions and (b) implicit knowledge of the MD-rule for proper names, he was already able to insert the proper name into discourse, even if only to find that he was mistaken.
(vii) A third counter-example suggested by Donnellan is person A, who wearing a special pair of glasses identifies two identical squares on a screen, placed one on top of the other. She calls the top square Alpha and the square on the bottom Beta. The only description suitable for identifying Alpha is its position. Now it turns out that unknown to person A the glasses invert the positions of the squares. Actually, Alpha is the square that appears on the bottom. Donnellan believes he has thus demonstrated that the square which A refers to as the Alpha square is in fact the square on the bottom, even associating it with the mistaken description (a): ‘the square to be seen on top.’
In response, I propose A only refers to the Alpha square, associating with it an erroneous description because it is a convergent description (since it correctly identifies a square in the correct spatial dimension). This description is correctable to:
(b) Alpha (Df) = the square that A sees as on top… even though it is in fact the square on the bottom, because A is wearing glasses that invert the positions of the images.
Although observer A does not know description (b), this description is the complete localizing description of the square Alpha, known by privileged speakers such as B. Thus, speaker B knows that square Alpha is on the bottom, for he has the information expressed in (b), which gives the referent’s complete mode of presentation. A has a convergent but incomplete and erroneously interpreted description. This is made clear because, once she is informed by B about the inversion of images by the glasses, she will soon replace description (a) with Alpha’s true identification rule, which includes (b).
10. Explanatory failure of the causal-historical view
Finally, let me say something about the causal-historical view. I do not wish to deny that there is some direct or indirect causal or causal-historical relation between the utterance of a name and the first tags of a name’s bearer. Even descriptivists like P. F. Strawson did not deny this. After all, we live in a world of causes and effects, and a proper referential link must have some causal dimension. What I reject is its explanatory relevance. No one uses it as a form of explanation. If someone asks me who Aristotle was, I do not answer: ‘All you need to do is to continue following my causal-historical chain, without forgetting to have the intention to refer to the same Aristotle I refer to.’ Indeed, in themselves the causal-historical links will remain inscrutable unless in searching for them we appeal to something like correlative cognitions and consequently to descriptions representing these cognitions. Suppose we had, for instance, a cerebroscope: with such a device we could show that whenever a speaker says the name ‘Aristotle’ and really knows whom he is speaking about, a recognizable neurophysiological pattern arises in his brain. We could identify this pattern as a link of the external causal-historical chain and search for similar links in other speakers. But since in this case we would need to appeal to the speaker’s cognitive-intention, implicitly we would be appealing to descriptions. This shows that the appeal to the causal-historical chain as an explanation commits a petitio principii by presupposing descriptivism. To make things worse, Kripke’s view of baptism is magical, since it cannot really be based on any property of the referent – it testifies to a form of referential mysticism that blocks paths of inquiry. Indeed, if we pick out some property, we will be having a thought or intention, and consequently this can be descriptively translated. Philosophically challenging, as is much of Kripke’ work, if taken at face value, the causal-historical view of proper names is nothing but a philosophical fantasy that begs the question. As someone once said, for any complex problem there is always a simple answer that is unequivocally deficient.
 This appendix is a short version of a much longer paper entitled ‘Outline of a Theory of Proper Names’ (Costa 2014a).
 I adapt this synoptic formula from Susan Haack’s interpretation (1978: 58).
 To be more precise, C must be the nearest most relevant class that does not merge with the characterizing description. This is why for the name Aristotle C must be the condition of being a human being and not of being a philosopher.
 I place the name in quotation marks to indicate that it must be possible to be misleading about the true symbolic form of a proper name. Imagine a possible world where only one philosopher satisfies the fundamental conditions for being our Aristotle, but who is called ‘Pitacus.’ We would after all still identify him with our Aristotle! Indeed, even in our real world we cannot completely exclude the possibility that our Aristotle was in fact called Pitacus… What individuates a proper name is the identification rule we associate with it. The description ‘the man named “Aristotle,”’ popular in the so-called metalinguistic theory of proper names, is only a well-known (accidental) auxiliary rule.
 In this connection, we could ask about the role of causality. Though I now believe that causality plays an important role in naming – we couldn’t have naming without any place for causality – it is usually presupposed. Imagine the causal role of a fortune-teller, who after gazing into a crystal ball always correctly guessed the name of unknown new visitors… If we believed in something beyond stage magic, we would assume some occult causal determination, e.g. clairvoyance. Suppose now that by an incredible coincidence I dream of a beautiful snow covered volcano called Ozorno and located in Southern Chile. This volcano really exists, which means that my reference is purely coincidental. But is a purely coincidental reference more than a mocking reference?
 Kripke also calls a rigid designator a term that designates the same object in every possible world (even in worlds where the object does not exist) (1980: 49). This would cover cases like that of the contra-factual supposition that Hitler was never born, in which case the name ‘Hitler’ would refer rigidly in a possible world where he never existed. However, I am unable to find any intuitive sense in such unnecessary conceptual contortionism.
 There are several decisive arguments to counter the old objection (Ziff 1960: 93-94) that proper names have no meaning. For one thing, identity sentences like ‘Mt. Everest is Chomolungma’ are informative, which means the names must have different senses. Moreover, proper names belong to classes designated by meaningful predicates, e.g., Mt. Everest is a mountain and not a prime number, and the existence of proper names’ references can be negated. Thus, the statement ‘Vulcan does not exist’ suggests that the name Vulcan must at least have a meaning, since the statement does not deny the existence of the word ‘Vulcan’ (cf. Searle 1969: 165 f.). My own view is that proper names have so much proper meaning and such diverse intended meanings that at first view they may appear to have no meaning at all.
 I am not the first to see the inadequacy of this counter-example. As one commentator wrote: ‘But why then say that he has in the appropriate sense referred to someone? The language we use to describe such cases may be misleading. Suppose, for example, that Jones is trying to remember whether he has invited someone besides A, B and C to dinner; he has the feeling that there may be one or two more. One might say to him: “Whom might you have in mind?” The point is that in other rather more central uses of these expressions, there is in this sort of case no one whom he has in mind, or means, or refers to; he cannot remember’ (Brian Loar 1976: 367).
 In the case of ‘the last living Neanderthal man,’ a one-foot fundamental description, what really counts is whether the characterizing description applies. The localizing description is only a statement about a space and time whose beginning is vague and whose end is unknown and can be different in other possible worlds.
 Note that this intention would have no proper cognitive content, otherwise we would be able to express this content linguistically, falling back into descriptivism. But without proper content, this intention would be nothing but a desire, a bet on the sameness of reference.
 The supposed semi-fictional name originally used as an example by Kripke was that of the biblical prophet Jonah (1980: 67-68). However, most serious Bible scholars believe that Jonah was in fact a purely fictional character.
 R. B. Marcus (1993:11) earlier proposed a similar example using the name ‘Venus’.
 Defending non-descriptive senses as mental files, François Recanati accepts the suggestion that in perception, an object can arguably be thought of without properties (Recanati 2012: 29-31). This is true if you expect to identify things like material objects or natural kinds. But when you think about very basic properties like density, hardness, volume, form, color, warmth… something must be present. Without properties like these no object can even be hinted at.