Forthcoming in the Proceedings of the Hawaii
International Conference on Arts and Humanities,
Honolulu, Hawaii, January 11-14th, 2003

Empty Names, Natural Kind Terms, and Radical Externalism
Frederick Adams
Laura Dietrich

I Introduction
There are many different names and kind terms.  Those that have bearers are filled names—names such as “Bill Clinton” or “George W. Bush,” “water,” “salt.”  Those that lack bearers are empty names—names such as “Santa”, “Vulcan,” or “phlogiston,” or “ghost.”  For direct reference theorists and radical externalists about kind terms, sentences employing filled names express propositions into which are inserted the objects or properties named.   For example, the sentence “Clinton was impeached” expresses the proposition identified by the ordered pair <Clinton, impeachment>.  Although there are many descriptions associated with Clinton, such as, “first president in the 21st Century,” “last president to be impeached by the House of Representatives,” and so on, none of these are part of the content of the proposition expressed solely by the name “Clinton.”  According to theories of direct reference, names contribute only their bearers to propositions expressed by their use.
 What propositions are expressed by “Pegasus flies,” “Vulcan is the tenth planet,” or “phlogiston is released during burning.”  Empty names or kind terms cannot contribute their bearers to propositions expressed by these sentences.  So what is the content of the sentences, and what are the meanings of the empty names or kind terms?  One answer available to radical externalists and theorists of direct reference is that empty names or kind terms lack meaning and propositions expressed by their use are incomplete or gappy.   So, for example, “Pegasus flies” expresses the incomplete proposition identified by <___, flies>, and “phlogiston is released” expresses <___, being released>.  These structured entities are incomplete propositions.  They have slots for things designated by subjects and predicates, but whether a slot is filled depends upon whether the name is filled or empty.  Furthermore, since the propositions are incomplete (or gappy), they are neither true nor false.  Consequently, on this view what is expressed by “Pegasus flies” or by “phlogiston is released during burning” are neither true nor false.  More controversially, the content of “Pegasus does not exist” is neither true nor false—and so on for other negative existential sentences employing empty names or kind terms.
 This consequence is highly counterintuitive, we realize.  Yet theories of radical externalism and direct reference have mechanisms to account for the contrary intuitions.    On the view that we accept all names and kind terms (empty or filled) have associated with them various descriptions.  None of these descriptions give the meanings of the names or kind terms.  Nonetheless, the descriptions become associated with names by learning.  These psychological associations can explain why names and kind terms (empty or filled) can seem to have meanings other than their bearers.   So for example, every U.S. schoolchild learns that George Washington chopped down a cherry tree and confessed and later became the first president of the U.S.  So we are pretty much conditioned (stimulus-response style) to connect “confessed to chopping down a cherry tree” and “first president of the U.S.” with “George Washington.”  Indeed, associated with all names there will be sets of descriptions (that we will call the lore ) that are associated with the name.
 There is lore associated with empty kind terms as well.  Associated with “phlogiston” are the descriptions “term introduced by George Stahl,” and “substance formerly believed to be released during burning.”    When we utter “phlogiston is released,” although we literally express the incomplete proposition <___, being released>, we pragmatically imply  complete propositions that would be expressed by taking a description associated with the name and substituting.  So, for example, we pragmatically imply that a substance believed by Stahl to be responsible for burning is released.  And if we utter “phlogiston does not exist,” we pragmatically imply that there is no substance named by Stahl released during burning.  We claim that this latter implied proposition is complete, true, and its truth misleadingly inclines us to think that a sentence such as “phlogiston does not exist” says something true.
 If we are right, there are at least two mechanisms at work: learned associations and pragmatic (as opposed to semantic) implications.  And there are at least two potentially misleading outcomes due to these.  First, one may come to think that one or more of the descriptions (“substance released by burning”) gives the meaning of a kind term (“phlogiston”), when the term has no meaning.  Second, one may come to think that one can express a truth using a vacuous term (“phlogiston does not exist”), whereas one at best pragmatically implicates a truth (that there is no substance named by Stahl).
 We are attracted to this theory for several reasons.  First, it offers a unified account of the meanings of names and kind terms.  It says that in all cases, the meaning of a name on an occasion of use is its bearer.  Mixed accounts could say that the meaning of a filled name is its bearer, but the meaning of an empty name is a description.  We are inclined to think that it would be preferable to say that names make the same type of contribution to what is expressed whether filled or empty.  Our account lets us say this.  We believe that one should move to the mixed account only if a unified account fails.  Part of our goal in this paper is to show that a unified account succeeds against several recent objections.  Second, the account applies to fictional names as well as non-fictional names (Adams, et.al. 1997).  Once again, one may propose a mixed view where the meanings of non-fictional names are their bearers and the meanings of fictional names are descriptions or characters, or some other entities.  We continue to believe that one should move to such a mixed view only if the unified view fails, and we will defend the unified view here.  We will now consider objections to this radical externalism that we have just outlined and its treatment of empty names and kind terms.
II Slim Support for Narrow Content
In his recent book, Gabriel Segal (2000) argues against externalists’ accounts of cognitive content and for a kind of narrow, individualistic content.  Segal defends a kind of content on which “being in a state with a specific cognitive content does not essentially involve standing in any real relation to anything external” (11).  And he defends a local supervenience thesis on which “ if two beings are identical in respect of their microstructural properties, then they must be identical in respect of their cognitive contents” (14).  Contrary to many narrow content theorists, he holds that narrow content is a variety of ordinary representation.  For example, in a Twin-Earth scenario  where Zowie’s “diamond” concept was formed on diamonds and Twin-Zowie’s “diamond” concept was formed on Twin-diamonds, Segal’s “view is that both Zowies’ diamond concepts apply to both diamonds and twin diamonds” (19).  If true, it seems to us that both Zowies have garden-variety broad, disjunctive content—not narrow cognitive content.  There is nothing narrow about concepts of diamonds (or Twin-diamonds).  So we have a hard time distinguishing Segal’s narrow content from what the tradition would call broad, disjunctive content.
We will not dwell on his positive account of narrow content here.  This is in part because he claims not to have a theory of content to offer (20), and in part because he says his “basic strategy is to undermine …the most popular and influential externalist theses and show that an internalist alternative is workable and attractive” (20).  We focus our attention on his arguments against externalism.  We believe his main argument against externalism is easily met.  If we are right that Segal’s negative argument against externalism is easily met, then his strategy of arguing indirectly for narrow content by exposing difficulties with broad content turns out to offer slim support indeed for narrow content.
III Empty Natural Kind Terms and Concepts
 Segal aims his attack at what he calls “the radical externalist position” as applied to natural kind terms and concepts.  This is the view that there is a world dependence of kind concepts upon their extensions.  In particular, the extension conditions of non-empty kind concepts depend upon a “real relationship between thinkers and samples” in the external world.  A thinker of kind K must have interacted with samples of kind K (or with someone who has).  In addition, this extension condition is essential to the cognitive content of the concept of kind K.  So if concept of kind K is world-dependent, then any concept that has a different extension condition than K has a different cognitive content from K.  This is intended to characterize atomic concepts whose content does not decompose into contents of parts (31).  Segal argues indirectly for kind concepts that have extension-independent cognitive content, by arguing: (1) that radical externalism cannot account for the existence of empty kind terms and concepts, and (2) that radical externalism cannot account for the sameness of intentional behavior of Twin-Earth twins, one of whom has an empty concept.   We intend to show how radical externalism can account for the existence of empty kind terms and concepts and as much sameness of intentional behavior as there is.
IV Segal’s Argument Sketch
Empty terms and concepts are the test case for the dispute between externalists and internalists about content.  Prior to Segal, Boghossian (1997) used his “Dry Earth” example to make a similar challenge to externalism—we discuss Boghossian’s arguments below.  And there have been other such attacks as well.  Similarly, empty names have been viewed as the test case for direct reference theories of proper names for many years .  So it is correct for Segal to look to empty kind terms and concepts as the front line in his attack on radical externalism.
 Segal presents the externalists with a dilemma in accounting for the mere existence of empty kind terms and concepts.  He asks what would happen to the meanings of kind terms and the concepts they express if the terms were vacuous.  He asks us to imagine Twin-Earth scenarios where terms such as “water,” “aluminum,” “topaz,” “quark,” and “polio” are empty on Twin-Earth but not empty on Earth.  Following Boghossian (1997) we can imagine that those on Twin-Earth do not know that the terms are empty (vacuous).  They may be systematically deluded or deceived or just mistaken about whether such terms refer.
 Segal maintains that the radical externalist must say either: (a) that such terms express no concept or (b) they express a concept “different form the one they actually express” (31).   We take the latter to be impossible, but what Segal may be saying is that they express a concept different from the one supposed by the user (thinker).
 Here is a parallel with empty names.   We would take Segal to maintain that for the externalist, “Santa Claus” either (a) has no content or (b) if it does express a content, it expresses a content other than that supposed by the speaker.  For example, “Santa lives at the North Pole” either expresses that “___ lives at the North Pole” or imparts but does not express that there is a jolly fat man who wears a red suit and brings presents on Christmas Eve and who lives at the North Pole.  In the first case, “Santa Claus” semantically expresses no content, though there is content expressed by the remainder of the sentence.  In the second case, the use of “Santa Claus” pragmatically imparts something, but does not express semantically what one would have expected, viz. that Santa Claus lives at the North Pole.   We will say more about this parallel as we go.  For we maintain that it provides the key to understanding empty natural kind terms and concepts.
V (i) No Content
Let’s start with the “no content” horn of the supposed dilemma for externalists.  Segal believes that the very existence of empty kind terms and concepts should be something of an embarrassment for externalists.   This we deny.  We agree with Segal that empty kind terms exist.  Many kinds of empty names exist.  Kind terms are a kind of name (in so far as they are a type of linguistic entity).  Semantically they are empty.  Syntactically, they may play the role of a common noun.  So there are indeed linguistic entities that play the role of natural kind terms and are intended by their users to refer to natural kinds (without success).
 Why say, as Segal does, that maintaining that empty natural kind terms have no content is “courageous?”  We suspect that it is because, for referentialists and externalists generally, there are going to be counter-intuitive results.  Such contrary intuitions have to be explained away.  We see this with direct reference theories and empty proper names no less than with externalism and empty natural kind terms.  So we maintain that, at bottom, Segal is exploiting the counterintuitiveness of externalism when he says “there are numerous empty kind terms that we must take to express concepts…they are pervasive and significant features of human cognition” (33).
 Here is an example of a corresponding counterintuitive result from empty proper names and direct reference theories of them.  Suppose one maintained that the very existence of empty names, such as “Vulcan,” “Santa,” “Sherlock Holmes,” is an embarrassment for theories of direct reference.   On such theories, one may hold that “Vulcan does not exist” does not express a truth because “Vulcan” is an empty name.  As such, it contributes no object to a proposition.  So when one employs a vacuous proper name in this negative existential sentence there is at most an incomplete proposition expressed of the form: there is not an x such that x = ___ (where one cannot fill the blank).  Hence, there is no truth (or falsehood) expressed using the term “Vulcan.”  This surely is counterintuitive and has led philosophers to say that such cases spell doom for theories of direct reference.  However, theories of direct reference explain away the recalcitrant intuition.  They explain it away by pointing to the lore (character, stereotype) associated with the term Vulcan consisting of descriptions such as “the tenth planet,” “the planet between Mercury and the sun,” “a planet causing perturbations in Mercury’s orbit.”  When one says “Vulcan does not exist” one literally expresses no truth, but one pragmatically imparts an associated proposition such as “that there is no tenth planet.”  The associated description, though not what is literally semantically expressed, is an imparted truth.  It is the truth of the pragmatically imparted information that explains a hearer’s (or speaker’s) taking a truth to have been uttered.   Hence, one is able to explain away the counterintuitiveness of the theory, and one is able to resist temptation to posit some extension-independent content for the term “Vulcan.”
V (ii) Content Other than Expressed
 Returning to natural kind terms and the other horn of the dilemma, we suggest a parallel account that explains away the counterintuitiveness of maintaining that empty kind terms semantically express no content.  Lore (character, stereotype) associated with the kind terms will have content and may be pragmatically expressed or accessed when employing empty kind terms.  Any cognitive content that appears to emanate from the empty kind term actually has its source in the associated terms and descriptions.  We will now turn our attention to some of Segal’s examples and explain how the “no content” ((a) above) view works.  At the same time, we will explain how this incorporates the view that a different kind of content other than what was expressed is the content that is imparted ((b) above).
 
 

V (iii) Examples
Bekong
At one point Segal introduces an empty kind term “bekong.”  This term derives from a group known as the Fang (a people who live mainly in Gabon and Cameroon).  The Fang believe bekong to be a type of ghost entity from which people can directly receive inspiration or messages and who can cause people illness or misfortune.  They believe the bekong to be the ghosts of their ancestors.
 Segal maintains that it is not an option to doubt that the Fang have a concept in the concept bekong (even if empty).  The concept plays an important role in the thoughts and behavior of the Fang.
 We maintain that the Fang’s term bekong is itself empty (on the assumption that there are no ghosts in Gabon and Cameroon).  And since the term is empty, there is literally no semantic content expressed in speech or thought involving this term.  We suspect that Segal would ask “how can you deny that the Fang have a concept bekong?”  There is a syntactic structure in their heads and it is causing thoughts and behavior, so how can you deny that they have a concept?
 Our reply is that they have a syntactic structure or name in thought and speech.  This name is playing a syntactic role.  But since it lacks a referent, it literally has no content itself.  However, it does play a causal role in the speech and behavior of the Fang, and the role it plays is explained by the cognitive content of the associated descriptions and lore (character, stereotype).  Since the associated descriptions have normal broad content, there is no difficulty explaining the role played by the empty bekong concept in terms of the non-empty content of the associated concepts and meanings.
 Segal’s description of the ways the Fang think of bekong is underdescribed, but we will provide some plausible elaboration.  Here are some reasonable guesses.  The Fang likely believe the bekong “are the sources of the voices they hear in their dreams.”  They likely believe the bekong “are the causes of the premature deaths of their infants.”  Perhaps they also believe the bekong “are invisible causes of disease, plague, and famine.”  So there are many descriptions that Fang associate with the term “bekong.”
 On a referentialist/externalist account of the meanings of names or natural kind terms these associated descriptions (lore, character, stereotype) do not constitute the meaning of a term for the usual Kripkean (Kripke, 1972) reasons.  But these descriptions can constitute the only contents associated with the empty terms.  So when a Fang utters a sentence such as “the bekong are angry,” the utterance literally expresses an incomplete proposition identified by the set consisting of a null subject and the property of being angry < ___, being angry>.  The sentence is neither true nor false.  And if an anthropologist studying the bekong utters “bekong don’t exist,” the anthropologist utters an incomplete proposition of the form that there is not an x such that x = ___ (where one cannot fill the blank).   The incomplete proposition is expressed using the term ‘bekong.’ This utterance too is neither true nor false.
 To explain the apparent truth of the negative existential sentence, we appeal to the information pragmatically imparted (but not literally expressed).  Imparted is information to the effect that there are not invisible beings who speak directly to the Fang in their dreams.  Or that there are not invisible conscious entities who cause disease, plague, and famine.  And so on.  Any apparent content expressed using the empty term “bekong” is accounted for by appeal to the content of the associated descriptions (lore, character, stereotype).
 This accounts for the mere existence of empty concepts on the radical externalist’s view.  We move now to the explanation of intention behavior.  Segal says “It is not a serious option to doubt that the Fang have an empty concept bekong.  It certainly appears that the concept plays an important role in their thought….beliefs, desires, and other cognitive states that motivate many of their activities” (35)  We agree that the Fang have an empty concept bekong.  We also agree that it plays an important role in the cognitive states and intentional activities of the Fang.  What we deny, and what Segal has not defended, is that concepts that do not themselves literally express a content cannot play these important cognitive roles.  On our view, the empty term “bekong” becomes psychologically associated with meaningful descriptions (lore, character, stereotype).  The cognitive role of the term “bekong” is explained via (1) its cognitive association with the meaningful descriptions and (2) the literal broad semantic contents of those associated descriptions.
 Let’s consider some examples.  First, consider a case where a thought about the bekong causes another thought.  Suppose the Fang want to appease the bekong and decide to best do so by giving them an offering.  So the fear of the bekong causes the thought that to appease the bekong it is good to give an offering.  The first cognitive state may be the fear expressed by the English sentence “The bekong are angry.”  This has the literal content identified with the ordered pair <____, being angry>.  It is expressed using the vacuous term in thought “bekong.”  How could this cause anything, if empty?  It could by being psychologically associated with several other thoughts from the bekong lore,  such as the thought that there are invisible conscious beings who speak directly to us in our dreams and they are angry.  It is associated with the thought that there are invisible conscious beings who cause premature deaths to our infants and those beings are angry.  These associated thoughts cause other thoughts about how to avoid the bad affects of the anger of these beings.  This leads to the intentions to appease these beings by giving an offering.  Since the incomplete thought “the bekong are angry” is cognitively linked to these other complete thoughts (even though they are false), it is quite easy to see how one thought (though empty) could cause other thoughts.
 As for the explanation of the intentional activities, it should be quite easy to see how these incomplete thoughts about the bekong can lead to complete thoughts and intentions about activities such as giving offerings (or other activities aimed at the non-existent bekong).  For these associated thoughts are complete thoughts (though false).  So we see no difficulty at all for the radical externalist who claims that empty concepts have no direct content.  Any cognitive explanations of thought or behavior is accounted for in terms of the perfectly contentful lore associated with the empty natural kind terms.  So we do not disagree with Segal that there are empty natural kind concepts.  We do disagree that their existence gives any good reason to believe in the existence of extension-independent (narrow) content.  Our account above makes no use of such content and does what Segal says cannot be done--provide a viable alternative to his view (38).

ME (myalgic encephalomyelitis)
In a second example, Segal gives a more traditional style Twin-Earthian argument for the existence of narrow content.  Earth Peter has the concept of ME (by its other name CFS chronic fatigue syndrome).  Twin-Peter has the same concept, but on Twin-Earth ME doesn’t exist.   From the twins’ epistemic perspectives, everything about symptoms of “ME” on Earth and Twin-Earth are identical (both are associated with dizziness, rashes, aches, sensitivity to light and sound, cold sores, swellings and forgetfulness).  Thus, the argument is that Peter and Twin-Peter must share some content to explain the sameness of their thoughts and behavior.  Since their concepts do not share extensions, there must be some extension-independent contents that their concepts share.
 Segal suggests that the assumption that Peter and Twin-Peter share the same concept of ME “would work perfectly well” (41).  Here he makes the standard assumption that the best explanation of the sameness of intentional behavior is that the twins share intentional content (in this case, the concept of ME).  He is appealing to the idea that the best explanation of why the concepts of the twins have the same cognitive causal roles is that they share the same content.  And “if the two concepts had only wide contents, then it is not at all obvious why their cognitive roles should be the same” (43).
 However, we have an explanation of why the roles would be the same even if their contents were different.  The reason is that both twins associate the exact same lore with the terms “ME,” even though on Earth the term is not empty and on Twin-Earth the term is empty.  The content of the associated lore is identical and shared by the twins.  The cognitive sameness of Earth Peter’s ME-concept and Twin-Peter’s ME-concept is to be explained in terms of the sameness of associated lore.  The sameness of causal role in thought and behavior is to be explained in the same way.
 So suppose Peter thinks “ME is making me dizzy, forgetful, and causing a rash.”  This thought causes Peter to go to the doctor for medication.  Peter’s thought expresses the proposition identified by the n-tuple <Peter, causation, ME, dizziness, forgetfulness, and rash>.  Peter associates with ME the descriptions “cause of dizziness, forgetfulness, and rash” (among other things).  Peter existentially generalizes that he has a single kind of disease that is causing his symptoms and, therefore, he should seek a doctor.   His thoughts are true and cause him to seek a doctor for help.
Twin Peter has a thought of the same form “ME is making me dizzy, forgetful, and causing a rash.”  Though he doesn’t know it, his thought expresses the incomplete proposition identified by the n-tuple <Twin-Peter, causation, ___, dizziness, forgetfulness, and rash>.  Twin-Peter’s thought lacks a truth value.  Since Twin-Peter does not know his thought is incomplete and lacks a truth value, he existentially generalizes that he has a single kind of disease that causes these symptoms and, therefore, he should seek a doctor.   His thought that he has a single kind of disease is false yet causes him to seek a doctor for help.  Twin-Peter has no disease ME, but he does have the same symptoms Peter associates with ME.   Twin-Peter’s thoughts that he has these symptoms are not false.  He, like Peter, believes he has a single disease called “ME” causing the symptoms.  He like Peter goes to the doctor for help because he has this thought.  Unlike Peter, his thought that he has ME is false.  Like Peter, his thought that he has these symptoms is true.
 Epistemically, since Twin-Peter is in exactly the same situation as Peter, his cognitive contents work the same in explaining the sameness of his (and Peter’s) behavior.   The differences in the broad contents of their thoughts factor out precisely because the differences are epistemically opaque to the agents.  The differences also factor out of the explanation of the sameness of their behavior.
 Therefore, we see no difficulty whatsoever for the radical externalist to explain the sameness of behavior that there is between Peter and Twin-Peter.  There is no need of extension-independent content to explain this sameness of the behavior of Twin-Earth and Earth twins.  Though the content of their concepts are different (one being empty, one not), the contents of their associated concepts are the same.  It is the contents of the associated concepts that share the same broad contents plus the cognitive association with the empty (and filled) concepts that explains the sameness of the twins’ behavior (what sameness there is).  Segal has the same strategy for explaining sameness of behavior of the twins as we do.  It is just that he wants to find shared content in the very concept of ME.  Whereas, we want to find only shared content in the associated concepts and descriptions.  We see no reason to prefer his shared content to ours.  Therefore, neither of these examples lends support to Segal’s arguments against radical externalism of concepts.
 We will return to two more of Segal’s arguments when we consider the case of Swampman (and supervenience) below.  Next we wish to establish that our reply to Segal also satisfactorily handles Boghossian’s “Dry Earth” example.
VI Dry Earth
In the context of arguing that there is an incompatibility between externalism and certain kinds of a priori knowledge of our thought contents, Boghossian (1997) introduces the example of Dry Earth.  We won’t here address the issue of compatibility with a priori knowledge.   Instead we will focus attention on Boghossian’s claim that there must be a kind of content incompatible with externalism.
 Boghossian introduced this as a case of an empty natural kind term and a powerful difficulty for externalism.  Boghossian’s (1997) Dry Earth is a Twin-Earth style example where Dry Earth is a “planet just like ours in which, although it very much seems to its inhabitants that there is a clear, tasteless and colorless liquid flowing in their rivers and taps to which they confidently take themselves to be applying the word ‘water,’ these appearances are systematically false and constitute a sort of pervasive collective mirage” (170).
 Boghossian employs the example in order to raise the potentially embarrassing question  “what concept, if any, should a Twin Earth externalist say would be expressed by tokens of the word ‘water’ on this Dry Earth? (170)”  Boghossian wants to say that on Dry Earth, ‘water’ expresses a concept even though it fails to refer. But what concept does it express?  He suggests that one could say that it expresses a compound concept such as: “the clear, tasteless, colorless liquid that flows in the taps and the rivers around here now….” (171).  Boghossian finds this suggestion to be plausible, but not avialable to the Twin Earth externalists because it would make ‘water’ an atomic concept when it referred, but a complex, non-atomic concept when it failed to refer (where the only difference was an external difference).
 Of course, we do not accept the compound description as a plausible candidate for the content of “water” on Dry Earth.  So we do not run afoul of the atomic/non-atomic problem.  For us the concept is atomic but empty.  Boghossian presses one who takes this line as follows:  “What proposition—what truth condition—is expressed by sentences of the form, ‘Water is wet’, for example, as uttered on Dry Earth?  What is it that gets said?” (173)
 Our answer, of course, is that “Water is wet” express the incomplete proposition identified with the order pair <___, being wet> (and expresses it using ‘water’).  This incomplete proposition is neither true nor false, but it is what gets said.  Of course, we also maintain that on Dry Earth one who utters this sentence may pragmatically impart the false proposition that there is a cool colorless liquid that flows in the taps and rivers around here.  This will be false, but will explain the apparent truth to Dry Earthlings and help explain their behavior (or the sameness of their behavior with Earthlings, such sameness as there is).
 Boghossian further claims that such a line of response by the externalist is fatal to the claim that there is a concept in the first place, but this would be like saying that an empty proper name ‘Vulcan’ is not a name.  True, it does not have a semantic value, but there is more to being a name than successfully naming.  Names have syntactic form and can form psychological associations.  Concepts do no less.  Both do this whether they are filled or empty.  So it is as much a mistake to claim that the externalist cannot admit the existence of empty concepts as to say that they cannot admit the existence of empty names.  Thus, we reject Boghossian’s claim that concepts must have non-empty extensions for externalists.  Natural kind concepts can exist and be empty even on externalist accounts.
VII Swampman
We have answered Segal’s and Boghossian’s challenges to radical externalism.  We will now conjoin two more arguments that Segal takes to support the existence of extension-independent content.  Both are types of argument that appeal to supervenience and both are susceptible to the now standard swampman type of challenge—which we will present and defend.
 Consider once again the “ME” example above.  Let Peter and Twin-Peter be physical duplicates.  According to the thesis of supervenience, mental properties supervene upon physical properties and there can’t be a mental difference without a corresponding physical difference.  There is no physical difference among the twins.  So, Segal maintains, there is no difference in conceptual content.  Peter’s concept ME must be identical to Twin-Peter’s concept ME (44).  Segal claims the physical neurological bases of the twins are identical and that the environmental differences of the twins are negligible.  True, Peter’s world includes the ME virus and Twin-Peter’s doesn’t.  Instead it includes a motley of factors that produce Twin-Peter’s symptoms.  But Segal adds “[b]y assumption, his concept is empty.  It has no extension.  It does not apply to the motley of phenomena….Everything present on [Earth] that could be relevant is also present on [Twin-Earth]” (45).  So Segal concludes that Peter and Twin-Peter share the same extension-independent concept ME.
 Before we discuss this further, Segal’s third argument adds to this only that for  nativists (such as Chomskians), one may wish to consider diachronic conditions of concept acquisition.  One would want to know how Peter and Twin-Peter acquired their concept of ME.  “In order to explain this, we would have to develop a complex theory of [their] innate endowment and developmental history…how [they] learned about the diseases, and so on.  But the key point is that whatever truth of the matter, everything specified by that theory would be present on [Earth and Twin-Earth]” (46).  So whatever conditions are sufficient for Peter to acquire the concept of ME are available for Twin-Peter to acquire the concept.  What this argument adds to the above is the innateness hypothesis.  If concepts are innate, then they can develop in different kinds of environments, making it more plausible that the twins (in different environments) share the same concept.  On the synchronic supervenience hypothesis, any physical difference in one environment could lead to a difference in mental content.  On the diachronic, sameness of innate content may tolerate some differences in physical environments.
 Our objection to these claims is that there is one crucial difference in the environments of the twins.  Peter causally interacts with the ME virus.  Twin-Peter does not.  This will cause a difference in content.  Peter’s concept is of ME.  Twin-Peter’s concept is of nothing (though it is psychologically associated with concepts of causes and symptomatic properties associated with “ME”).
 Segal anticipates this reply and rejects it on the grounds that it is implausible because “developmental psychologists do not usually care whether the concepts they study are empty or not” (47).  We find this to be irrelevant since developmental psychologists do not usually discuss Twin-Earth style cases (nor do they usually care about a philosophical theory of semantic content).
 The differences in content of the twins’ concepts is not something that is epistemically accessible to the twins (or to developmental psychologists).  So there is no reason to suppose that the twins would notice the difference (or that developmental psychologists would—uninformed as they are about the differences on the twin planets).  Since the concepts associated with ME are the same for the twins, the association of these same concepts can explain the sameness that exists for the twins’ thoughts and behavior.  That sameness would be sufficient for any developmental psychologist (or social scientist).  However, the difference in causal relations to one’s enviornment for the twins is crucial to the difference in content.  One cannot think about a thing or property without interacting with the thing or property (or with one who has)—or so we maintain.
 The swampman examples are the crucial test cases for the dispute we are now having with Segal.  For these examples test our (and his) intuitions about just this issue: must one interact with a property or its instances to have thoughts the contents of which are that very property?  Segal says no.  We say yes.
 Along with twin cases, the test cases for theories of content have always included brains in vats or physical duplicates with no prior causal histories with an environment.  Such is the example of Donald Davidson’s (1987) swampman.  Let Swamp-Peter be a physical duplicate of Peter (as is Twin-Peter), but with no prior causal history.  Davidson has him fly together out of particles after lightning hits a stump in a swamp (hence, “swampman”).  If Segal’s appeals to supervenience are correct, then Swamp-Peter should have the same concept ME as Peter (and Twin-Peter).  Segal admits that a claim such as that Swamp-Peter has the concept ME is “less plausible than its analogue [for Twin-Peter]….because [Swamp-Peter] has nothing resembling a history of acquisition of the concept” (52).
 Segal points out that if one assumes Swamp-Peter is subject to psychological explanations, then it is easier to maintain that Swamp-Peter should be said to share the same ME concept with Peter and Twin-Peter.  However, we see no good reason to maintain initially—in the few hours after Swamp-Peter materializes—that his movements are subject to psychological explanation.   Just because there are syntactic structures in Swamp-Peter’s head that are causing his bodily movements, that does not mean that his bodily movements constitute intentional behavior.
 Segal goes on to maintain that it would be “unprincipled” for the externalist to deny that Swamp-Peter has a concept of ME.   But his claim is based upon one’s accepting his arguments that have gone before.  Since we have systematically rejected those arguments, and provided principles for doing so, our denial here is hardly “unprincipled.”  If Swamp-Peter utters the sounds “ME hurts,” we maintain that he has uttered nothing meaningful to him, nor has he expressed a proposition or a thought.  This is not because he hasn’t interacted with ME but because he has no intentional states whatsoever.
 Segal’s last gasp is the charge that to deny Swamp-Peter any thoughts or a concept of ME is “counterintuitive” (55).  If true, so be it.  We are in the business of defending radical externalism.  Our account of empty proper names and empty kind terms also requires that we hold negative existential sentences such as “Santa doesn’t exist” or “phlogiston does not exist” to lack truth values (but be associated with imparted truths).  We are willing to give up an intuition here or there to preserve a theory that us unified and treats all names (proper or common) the same.  So we are willing to embrace the counterintuitive result of maintaining that Swamp-Peter has no concept of ME.
VIII Descriptive names and kind terms (McGinn)
There are two types of reply that one may make to the defense of radical externalism that we have given.  One (due to McGinn, 1989) is that that content of an empty kind term could be complex and non-atomic and may not depend upon real relations to instances of the kind.  The other is based on the possibility that empty kind terms may be treated as descriptive names.  We think neither of these replies work.
 Consider the first suggestion.  McGinn claims (35) that one may have the concept of H2O without having interacted with instances of water.  One may have the concept of H2O, having interacted with H and with O (but not their compound substance).  Notice that McGinn did not say that one who has this concept will thereby have the concept of water.  For water may be an atomic concept, and it does not follow that if one has the complex concept H2O, then one has the atomic concept water.  Even though these concepts are of the same stuff, they are not necessarily the same concepts.  To have the concept of water, one need not have the concept of H (but not so for the latter).
 While we accept that one may have the complex concept H2O without interacting with its instances, we are less sanguine about McGinn’s claim that one may have the concept coal by composing the complex concept “wood compressed in a certain way over millions of years.”  One need not have the concept of wood to have the concept of coal (in the days before we discovered its chemical nature).
 Even setting aside this worry, McGinn’s examples are cases of property identities.  They will not rescue the examples of Boghossian or Segal.  In the Dry Earth example, Boghossian is not claiming, nor should he, that water = the clear, colorless, odorless stuff that flows in the rivers and taps around here.  Water may be in a gaseous or solid state.  In the Segal examples, it is not being claimed that ME = the cause of a set of symptoms.  The ME virus could cause different symptoms under different circumstances.  And the same is the case with the bekong.  The property of being a bekong is not being identified with a set of contingently associated descriptions the Fang use to think about them.
 So even if one could think about H2O without interacting with it, it does not follow that the Fang can think about bekong, nor that Twin-Peter can think about ME.  For the descriptions associated with these entities are not descriptions of property identities.  The same can be said about Boghossian’s example.  Being water is not identical with its phenomenal manifestations of the associated descriptions.  So thinking of the latter does not constitute thinking of the former (even setting aside the issue of atomic vs. non-atomic disparities).
 Hence, while McGinn may well be right about the possibility of having complex concepts without interacting with their instances, this won’t rescue the examples we’ve discussed and rejected as problems for radical externalism.
 Turning to the second possible reply, one may treat empty concepts as a type of descriptive name.  One might maintain that all empty names are descriptive names.   We will now explain why one might be tempted toward such a view and why we think it fails.
Kripke (1972) and Evans (1982) discuss the possibility of descriptive names.  An example of a descriptive name is introducing the name “Julius” to stand for whoever invented the zip.  Both Kripke and Evans suggest that names can be introduced in this way.  If someone, say Ken, actually invented the zip then “Julius” refers to Ken and can be used to make assertions about Ken (“That Julius was a smart guy”).
 Evans (1982) adds the remark that descriptive names are names “even if the name is empty” (31).  Evans may go on to say that if no single person invented the zip, then the name fails to refer.  What then of the negative existential sentence “Julius does not exist?”  Since the name was introduced via the description “the inventor of the zip” and there is no referent, the content of this utterance may be taken to be that the inventor of the zip does not exist--which would be true.  So on this view, empty names, if descriptive names, would not express gappy or incomplete propositions.  “Julius is tall” would express the proposition that the inventor of the zip is tall (and would be false).  And so on.  Hence, if all empty names are descriptive names, one may be able to avoid the unintuitive consequences our theory is required to explain away by appeal to the mechanisms of association and pragmatic implication.
 One may be tempted to adopt this view of empty concepts for similar reasons.  It would be intuitively acceptable, and it would straightforwardly make negative existential sentences true assertions (not pragmatically implied truths).  For example, in the case of the Fang’s bekong, one may say the empty concept is introduced via the natural kind term and the associated descriptions.  If there is something that satisfies the associated descriptions, then the term “bekong” applies to that thing.  If there is nothing that satisfies the descriptions (as we assume to be the case), then the term fails to refer and the concept is empty.  However, to say “the bekong are angry” is to say “the Gs are angry” (where “the Gs” are the associated descriptions) and the sentence is false.  And to say “the bekong do not exist” is to say that there is nothing that is G (which is true).
Similar treatments are available for Segal’s example of “ME” and Boghossian’s Dry Earth.  For example, in the latter, the term “water” would be introduced on Dry Earth via the description G (= the cool, clear, colorless liquid…).  If there is something G, then “water” applies to it.  “Water is wet” would express that the G is wet.  “There is no water” would express that there is nothing that is G.  And so on.
So this way of handling empty kind terms has much to recommend it.  It avoids the unintuitive consequences of the view we are defending here—a view that utilizes incomplete propositions and truth-value gaps, and pragmatic implications.  Making empty kind terms descriptive names avoids all of that.
 We hanker for an account in which all names and kind terms make the same semantic contribution to their informational (or propositional) content on occasion of use.  We shall explain why the theory of names and kind terms we just described is not a unified theory of names.  On it, filled names and descriptive names have two different semantic functions (make two different types of informational contribution to propositional content on occasions of use).  The remainder of this paper will be devoted to explaining this and defending our account for the reason that it alone offers a unified theory of the semantics of names and kind terms.  Until or unless the unified account is shown to have a fatal flaw (or we are overwhelmed by the beauty of an alternate theory) we maintain that its simplicity and economy are still among its major strengths.
 Descriptive names and kind terms can be filled or empty.  Let’s first consider cases where they are filled.  So let’s suppose that Earthlings first introduced the name “ME” for the single unique kind of cause of the symptoms associated with this illness (where “this” demonstrates those with the disease, and the symptoms are those we listed above).  Let this description = G.  So “ME” is introduced via G.  Then suppose a unique kind of virus V on Earth causes the disease.  Then since V is G, “ME” names virus V.  So if Peter has ME, Peter has virus V.
The unique thing about descriptive names and kind terms is that the route from the introduction of the name to the bearer is not the same as Kripke’s “initial baptisms,” for proper names, nor direct causal contact for kind concepts.  Normally, when someone introduces a name there is a more or less direct causal chain between the introduction of the name and the bearer of the name (causal via the cognitive capacities of the one introducing the name).  So, at birth and in her presence Laura’s parents decided to name her “Laura.”  In the case of descriptive names, there may be no direct causal path from the introducer of the name “ME” to the virus V.  The description associated with the name G, (the lore, if you will) is a recipe for connecting the name to its bearer.  Since virus V fits the description G (the lore is true of it)--it is via the lore G that “ME” refers to virus V.
 Okay.  So far so good, but now when one utters “ME is a draining disease,” what proposition does one express?  There are two choices:
(a) The G is a draining disease,
(b) ME (aka Virus V) is a draining disease.
Some proponents (Evans, 1982, Recanati, 1993) of descriptive names would maintain that (a) is correct.  We maintain that only (b) is correct.   We maintain this because we maintain that names always make the same semantic contribution.  They contribute their bearers (if they have bearers) to the proposition they express.  It is true (in the example) that ME (aka virus V) is the G.  However, if the name “ME” made the semantic contribution the G on an occasion of use, then the meaning of the name “ME” would be the description “the G.” But the meanings of names are their bearers (if they have bearers).  Or at least, that is the theory we are defending.  So a name might be introduced via a description (a piece of lore associated with the name). Even though the description might help link the name with its bearer, once linked (and, hence filled) the informational content of the name on an occasion of use is its bearer.  There is no reason to maintain that the meaning of the name is other than its bearer.
 Why do some philosophers believe otherwise?  We maintain that it is because they mistake the pragmatic implicature associated with a name via the lore for the semantic content of the expression using the name.  So when one claims that (a) is true, we would agree, but argue that it is true only because it is a pragmatic implicature.  (b) is the literal semantic content of “ME is draining.”
 We would add that even Recanati agrees that once the name “ME” becomes established as a name for virus V, and we have additional lore associated with “ME”, the description “the G” may lose privileged status.  We may even come to learn that it is true that ME is not the G (or is G no longer).  This could not happen in any possible world where “ME” had “the G” as its meaning, but could happen where “ME” meant ME (aka virus V).
 So that is what we would say about filled descriptive names and kind terms.  There are no surprises here.  Our account says the same thing about all names, filled or empty, descriptive or not.  They make the same kind of semantic contribution.  They are associated with various amounts of lore (or descriptions).  And they contribute only their bearers to the propositions they express on occasions of use.
 Let’s turn to empty descriptive names and kind terms.  What does our account say about them?  The answer should by now be obvious. Empty descriptive names or kind terms occur when a name is associated with a set of descriptions (or lore) and no object is the bearer of the name.  So suppose “ME” fails to refer on Twin-Earth because there was no virus V or the syndrome was caused by a motley of causes (as per Segal’s suggestion).  Then the question becomes what does “ME is draining” express?  On our account it expresses the gappy proposition identified by <____, being draining> and it does this by using the name “ME.”  It does not express the proposition  (a) that the G is draining because, while this is a true pragmatic implicature, the description “the G” is not the meaning of the name “ME.”  We would say what “ME does not exist” does express, but all readers know that by now.
 While we understand the motivation for thinking that consideration of descriptive names introduces a new theory of empty names, we have maintained that it does so at the expense of a unified theory of names.  The resultant theory would treat the meanings of names as descriptions, which we maintain they are not.  On our view, the descriptive part of descriptive names or kind terms (filled or empty) is just the lore associated with names or kind terms.
IX CONCLUSION
In this paper we defend the view that empty names and kind terms have no meaning because they have no bearers.  Having no bearers, they cannot contribute an object or a property to the content of a proposition expressed on an occasion of use.  Hence, they cannot be used to say or think things that are true or false.
 We have given a mechanism to explain away any appearances that empty names or kind terms can express truths in such cases as negative existential sentences.  We explain away the appearance of truth by virtue of the associated lore of empty names and the pragmatic implicatures generated by their use.
 In addition, we have considered several objections.  The objections of Segal and Boghossian focus not on empty proper names, but empty kind terms.  Their arguments and examples challenge our version of externalism to explain not only what sentences employing empty terms express, but also how empty concepts can cause thoughts and intentional behavior.  We explain how the contents of associated descriptions plus the syntactic objects that constitute tokens of empty terms in thought are up to these tasks.
 In addition, we consider two further challenges to our view based upon the idea that empty kind terms may be non-atomic and the challenge that empty kind terms may be kinds of descriptive names.  We explain why both of these challenges may seem promising, but fail to provide a better alternative than our own view.
 Thus, we offer our radical externalist account of empty names and kind terms as an interesting and plausible contender in the semantics of names and kind terms.  It is a unified view.  On it, all names make the same kind of semantic contribution to thought or speech.  Names of any kind contribute their bearers (objects or properties) to what is said or thought on an occasion of use (or thought).  For atomic names and kind terms, empty terms have no meaning and contribute no object or property to a proposition expressed.  However, associated lore can account for any apparent content or truth or falsity.
 

References

Adams, F. & Dietrich, L. ms. What’s in a(n empty) Name?

Adams, F., Drebushenko, D., Fuller, G., & Stecker, R. 1990: Narrow Content: Fodor’s Folly, Mind & Language, 5, 213-229.

Adams, F. & Stecker, R., & Fuller, G. 1992.  The Semantics of Thought.  Pacific Philosophical Quarterly, 73, 375-389.

Adams, F., Fuller, G., & Stecker, R. 1993.  Thoughts Without Objects.  Mind & Language, 8, 90-104.
Adams, F. & Stecker, R. 1994.  Vacuous Singular Terms.  Mind and Language, 9, 387-401.

Adams, F., Fuller, G., & Stecker, R. 1997.  The Semantics of Fiction.  Pacific Philosophical Quarterly, 78, 128-148.

Adams, F. & Aizawa, K. 1997.  Rock beats scissors: historicalism fights back, Analysis, 57, 273-281.

Adams, F., Drebushenko, D., Fuller, G., & Stecker, R. 1990.  Narrow Content: Fodor’s Folly.  Mind & Language, 5, 213-229.

Boghossian, P. 1997.  What the Externalist Can Know A Priori.  Proceedings of the Aristotelian Society, 97, 161-175.

Braun, D. 1993.  Empty Names.  Nous, 27, 449-469.

Davidson, D. 1987. Knowing One’s Own Mind.  Proceedings and Addresses of the American Philosophical Association, 60, 441-458.

Devitt, M. 1989.  Against Direct Reference. Midwest Studies, 14, 206-240.

Dretske, F. 1988. Explaining Behavior.  Cambridge, MA.: MIT/Bradford.

Dretske, F. 1995.  Naturalizing the Mind.  Cambridge, MA: MIT/Bradford.

Evans, G. 1982.  The Varieties of Reference.  Oxford: Oxford University Press.

Everett, A.  Forthcoming. Empty Names and ‘Gappy’ Propositions.  Philosophical Studies.

Grice, H. 1989.  Studies in the Ways of Words.  Cambridge, MA: Harvard University Press.

Kripke, S. 1972.  Naming and Necessity.  In Davidson D. & Harman, G.,(eds.), Semantics and Natural Language.  Dordrecht: Reidel.

McGinn, C. 1989.  Mental Content.  Oxford: Blackwell.

McLaughlin, B. & Tye, M. 1998.  Is Content-Externalism Compatible with Privileged Access?  Philosophical Studies, 107, 349-380.

Patterson, S. 1988.  Individualism and Semantic Development.  Philosophy of Science, 58, 15-35.

Recanati, F. 1993.  Direct Reference: From Language to Thought.  Oxford: Blackwell.

Segal, G. 2000.  A Slim book about Narrow Content. Cambridge, MA.: MIT/Bradford.