1
Descriptivism
1.1 Descriptivism Expounded
Some linguistic expressions serve the purpose of referring to objects outside of language. For instance, the proper names āDavid Cameronā, āWest Portā and āEdinburghā refer to the Prime Minister, my favourite pub and the Scottish capital city, respectively.1 As these terms purport to pick out a single object, they are singular referring terms. Individuals, pubs and cities are concrete, tangible objects. Other singular referring terms pick out events or even abstract objects. Think of āTour de Franceā or ā1 metreā. Definite descriptions of the form āthe Fā might also naturally be thought of as singular referring devices. For instance, āthe current US presidentā picks out Barack Obama. In contrast, the general term ālemonā refers to all and only fruit of the kind Citrus limon, and the general term ātigerā refers to all and only cats of the kind Panthera tigris. As these terms purport to pick out more than one object, they are plural referring terms. Such terms have extensions, which are the sets of objects to which they correctly apply. Specifically, ālemonā and ātigerā are natural kind terms, which have as their extension all and only instances or members of an underlying physical, chemical or biological kind.2 Such kinds are more or less mind-independent demarcations that group objects together at various levels of generality. For instance, my mother is a member of the increasingly more general natural kinds Homo sapiens, mammal and living creature. Arguably, if an object belongs to a natural kind, then such membership is an essential property of that object. My mother could undergo various changes pertaining to her physical appearance or psychological profile, but she could not possibly cease to be a human being. However, to use a metaphor, not all plural referring terms carve nature at her joints in the way natural kind terms do. Non-natural kind terms pick out instances of artefactual kinds. Thus āsofaā picks out long upholstered seats with a back and arms, and ācarburettorā picks out any device that blends air and fuel for an internal combustion engine. In these cases, there are no common, underlying scientific kinds, e.g. some sofas are made of metal and wool, others of oak and leather. What matters for something to count as a sofa or a carburettor is not its physical constitution, but whether it fits the job description or plays a certain functional role characteristic of sofas or carburettors.
Having distinguished these kinds of referring terms, let us now turn to their semantics. When it comes to referring terms, semantics has to do with those aspects of meaning that are relevant for their reference. Frege (1964/1893) thought that every linguistic expression possessed a reference of a kind appropriate to its category. For instance, declarative sentences refer to their truth-values, which he called āthe Trueā and āthe Falseā.3 That is to say, such sentences are sentential names of the abstract entities the True and the False. And Frege took predicates to refer to concepts, which on his view are functions whose value, for every object as argument, is a truth-value. There is no need to follow Frege here. Some types of linguistic expressions are intuitively not referential. Predicates have extensions, e.g. all red things fall within the extension of ābeing redā, and sentences have truth-values, e.g. āBen Nevis is 1,344 metresā is true. The semantics of predicates or sentences thus pertains to those aspects of meaning that are relevant for their extension or truth-value, respectively.
If we ignore Fregeās quirky terminology and instead take the most basic semantic value to be the truth-value of a sentence, then we can understand semantics as the theory of how the truth or falsity of a sentence is determined by the semantic values of the expressions that compose that sentence. In other words, the semantic value of an expression consists in the contribution it makes to the truth or falsity of a sentence in which it occurs. The semantic value of a singular referring term is naturally identified with its reference, and the semantic value of a predicate is naturally identified with its extension.
What then does reference consist in? Reference is a unique relation between a singular or general term and one or more extralinguistic objects. Reference is the most direct way for language to latch onto reality. Indeed the reference of a complex expression is determined by the reference of its constituents and their mode of composition. Reference is subject to the following compositionality principle:
(ComRef) | The truth-value of a sentence is determined by the reference (or semantic value) of its component parts and the way in which they are combined. |
The last clause is important since it matters how the very same component parts are conjoined. The sentence āThomas hit Jamesā may be true, while the sentence āJames hit Thomasā is false. It follows from (ComRef) that the truth-value of a sentence will remain invariant if co-referential component terms are substituted while everything else stays the same. Terms that have the same reference are interchangeable without change in truth-value. For instance, since the sentence āBono is the lead singer of U2ā is true, and Bono also goes under the name āPaul David Hewsonā, the sentence āPaul David Hewson is the lead singer of U2ā is guaranteed also to be true. Here is a substitution principle for reference:
(SubRef) | If the sentence āa is Fā is true and a = b, then the sentence āb is Fā is also true. |
We can say that reference is a property that a term possesses only if it can be used in certain sentences that have a truth-value. To use Fregeās example, given that the positive, non-fictive sentence āWilliam Tell shot an apple off his sonās headā lacks a truth-value, the term āWilliam Tellā lacks a reference. Of course predicates can also be used in sentences that have a truth-value, e.g. āis redā as in the false sentence āBuckingham Palace is redā. If we wish to depart from Fregeās framework when it comes to predicates, the fact that a term can be used in a sentence that has a truth-value should not be sufficient for it to count as possessing a reference.
What then is the meaning of a referring term? By āmeaningā we mean the semantic content of a term, i.e. that aspect of its overarching meaning which is relevant for determining the referent, if any, of that term. Meaning is what determines reference. Some aspects of this broader notion of meaning are semantically irrelevant in this way. The name āBonoā might give rise to attitudes of artistic approval which are absent in the case of āPaul David Hewsonā. One historically influential view is descriptivism, which has been defended in various versions by Russell (1994/1905), Strawson (1950), Searle (1958), Dummett (1978: Ch. 9), Kroon (1987), Lewis (1972, 1984, 1997), Jackson (1998a, 2004) and others; indeed its roots can be traced back to Frege (1994a/1892). The idea is that the meaning of a referring term is descriptive in nature. More precisely, the meaning of a referring term āaā is given by a set of definite descriptions: āthe Fā, āthe Gā, āthe Hā, etc. For instance, the meaning of āAristotleā is something like: the famous philosopher of antiquity, the teacher of Alexander the Great, the author of Nicomachean Ethics, etc. According to descriptivism, the descriptive content of a referring term āaā plays two distinct roles: it is what a competent speaker S knows when she understands āaā, and it is what determines the reference of āaā. Descriptivism is both a theory of meaning and a theory of reference. On the face of it, this view seems to blur the distinction between the first-order question of what the meaning and reference of a term is from the second-order question of what makes it the case that a term has the meaning and reference that it does. To use Stalnakerās terminology (1997: 535ā38), the first-order question is answered by a descriptive semantics, and the second-order question is answered by a foundational semantics. In taking descriptive content to be the meaning of a referring term and also to be what makes that term have the reference it has, it looks as if the question āwhat is the semantics for the term?ā is conflated with the question āwhat makes it the case that that term in our language has this semantics?ā However, descriptivists such as Jackson (1998a) by no means reject that distinction. They only insist that descriptive content can play a role in answering both questions.
Letās begin with the theory of meaning. When S understands the proper name āAristotleā, she knows its meaning. That name is synonymous with the descriptions āthe famous philosopher of antiquityā, āthe teacher of Alexander the Greatā, āthe author of Nicomachean Ethicsā, etc., and so those are what S knows in virtue of understanding the name. That is to say, S understands the name āAristotleā if and only if S knows that Aristotle was the famous philosopher of antiquity who taught Alexander the Great and wrote Nicomachean Ethics. And the kind of knowledge S has is a priori ā knowledge that is independent of any empirical evidence. If S is competent with āAristotleā, then she knows a priori that, say, Aristotle taught Alexander the Great. The speaker S could not grasp the meaning of āAristotleā while it is an open question for S whether Aristotle taught Alexander the Great. For what makes the name have the meaning it has is the fact that competent speakers mentally associate these descriptions with it. And if grasping the meaning of āAristotleā involves associating the description āthe teacher of Alexander the Greatā with that name, understanding the name suffices for knowing that Aristotle is the teacher of Alexander the Great. In short, following descriptivism, meaning is fully determined by competent speakersā mental associations. On this view, meaning is firmly in the mind of competent speakers.
But descriptivism is also a theory of reference in that descriptive content is what determines reference: a particular object is the referent of a referring term if and only if that object satisfies all the associated descriptions. These individually necessary and jointly sufficient descriptions express descriptive properties, e.g. āthe famous philosopher of antiquityā expresses the property of being the famous philosopher of antiquity.4 So, put differently, what determines whether an object is the referent of a term is whether that object instantiates all those descriptive properties, e.g. what determines whether someone is the referent of āAristotleā is whether he was the famous philosopher of antiquity, taught Alexander the Great, authored Nicomachean Ethics, etc. Assuming a definite description āthe Fā is an expression that refers to a unique object, if no object at all is F, or if two or more objects are F, then āthe Fā fails to refer. This poses a problem about uniqueness for descriptivism. As Plato was also a famous philosopher of antiquity, there is no single famous philosopher of antiquity. But on this view āAristotleā refers to an object just in case it uniquely satisfies all these descriptions, including āthe famous philosopher of antiquityā. So, it looks like āAristotleā turns out to lack a reference. One solution is to think of the set of definite descriptions as a cluster in the sense that āAristotleā refers to an object just in case it satisfies most of these descriptions. That way āAristotleā will refer to Aristotle even if one of the associated definite descriptions fails to uniquely pick out Aristotle, or even if one such description uniquely picks out someone else. This cluster version of descriptivism was invented by Searle (1958), but has subsequently been developed by Lewis (1972, 1984) who took the cluster to comprise a disjunction of conjunctions of most of the descriptions. So, to simplify matters, suppose the three descriptions āthe Fā, āthe Gā and āthe Hā constitute the cluster assigned to the name āaā. Then an object is the referent of āaā just in case that object satisfies either āthe Fā and āthe Gā, or āthe Fā and āthe Hā, or āthe Gā and āthe Hā.
We have so far been using proper names as examples, but descriptivism is also a view about general terms. Take the natural kind term āwaterā. On this view, the meaning of āwaterā consists in descriptive content which is both what competent speakers know when they understand the term and what determines its reference. The term āwaterā is synonymous with a definite description of the form: āthe clear, potable, tasteless liquid that fills the oceans and falls from the sky, etc.ā In the following āthe watery stuffā will be used instead as an abbreviation, where āwateryā captures all these superficial, readily observable properties of water. So, a competent speaker S understands āwaterā if and only if she knows that water is the watery stuff. As S is associating this complex description with āwaterā, she can know a priori that water has these watery properties. Moreover, following descriptivism, āwaterā will pick out anything that satisfies āthe watery stuffā, or at least anything satisfying enough of the descriptions encapsulated by that phrase. In other words, anything instantiating all or enough of the descriptive properties expressed by these descriptions will be picked out by āwaterā, and so count as water. For example, on Earth H2O satisfies most if not all of these descriptions, and so H2O is what the inhabitants on Earth refer to when they use āwaterā in various sentences.
1.2 The Identity Argument
In Section 1.1 we presented descriptivism but adduced nothing in its support. In Sections 1.2 and 1.3 we present arguments that supposedly buttress the descriptivist outlook. Why do we need meaning over and above reference, and in particular why do we need descriptive content? Can we not just identify meaning with what we in Section 1.1 called āsemantic valueā? In particular, whatās wrong with Millās view (1963/1843) that the meaning of a referring term is nothing but its reference? In Fregeās famous article (1994a/1892) āOn Sense and Referenceā, he made a crucial distinction between sense (German āSinnā) and reference (German āBedeutungā). Frege thought that unless referring terms were associated with distinct senses, we could not explain what it is that a speaker S knows when she competently understands these terms. In contrast, according to referentialism, meaning is nothing over and above reference.5 On this view, to know the meaning of a referring term is to know its reference. In this work (1994a/ 1892) Frege advanced his identity argument, which is targeted at referentialism. It aims to show that to know the reference of a term is to know more than is involved in knowing its sense. The identity argument is presented as a challenge to explain how true identity statements can be informative. Consider:
- (1) Hesperus is Hesperus.
- (2) Hesperus is Phosphorus.6
The propositions expressed by the sentences in (1) and (2) intuitively differ in how informative they are. Propositions, remember, are the semantic contents that sentences express. While (1) is trivial, (2) is an interesting piece of information. It is perfectly possible for a rational speaker to grasp both proposition...