Putnam on Reference

David Banach


The main task of this selection will be to critically examine the views on meaning and natural kinds and the arguments given for those views given by Hilary Putnam in "The Meaning of 'Meaning'" (1975). As an introduction to this, however, I would like to consider first some of the views presented by Putnam in  two earlier articles, "Is Semantics Possible?" (1970) and "Explanation and Reference" (1973). This will make it easier to understand the more complex discussion that follows.

                In "Is Semantics Possible?", Putnam begins by distinguishing a group of general names which are associated with natural kinds. Natural kinds are classes of things we regard as of explanatory importance and whose normal distinguishing characteristics are held together or explained by some deep-lying mechanisms or essential nature. (Putnam 1970, p. 139) That is, they are a recognized class of objects whose macroscopic, easily recognized properties are the result of some essential nature, possibly microscopic properties, that the class has in common. Putnam goes on to argue that the traditional theories of meaning cannot account for the properties of this group of general terms. The traditional view of meaning held that the meaning of a common noun is given by specifying a conjunction of properties. For any one of these properties, P, used to give the meaning of a natural kind term, X, the statement, "X's have P.", is an analytic statement. Also, if S is the set of properties used to give the meaning of X, then the statement, "Anything with all the properties in S is an X." is an analytic statement. This means that to say something is a member a natural kind is just to ascribe a conjunction of properties to it. Putnam argues that this cannot be true for a number of reasons: (1) A natural kind may have abnormal members that do not have all of the defining properties used to give the meaning of the natural kind term. (2) The defining characteristics of a natural kind may change without the essential nature of the kind changing. For example, the color of all lemons may change to orange, but they would still be lemons. It cannot be analytic that a member of the kind has these properties. (3) The essential nature of a natural kind is a matter for scientific theory to discover, not linguistic analysis. The defining characteristics we use to give the meaning of a natural kind term are dependent on our scientific theory. The theory may be mistaken in part or in whole. For example, if all the cats on earth turned out to be elaborate machines from Mars, there would still be a natural kind called cats, but it would not have most of the properties our best theories of today use to give the meaning of 'cat'. In general, it is impossible to determine in any mechanical way what things belong to a natural kind using a list of properties. Our knowledge of the essential nature of a natural kind and our knowledge of the connection between our list of properties and this essential nature are a function of our theories. It cannot be analytic that the natural kind has these properties. (Putnam 1970, pp. 139-43)

                Putnam goes on to suggest three conditions for knowing the meaning of a natural kind or term: (1) One must have implicit knowledge of certain semantic and syntactic markers associated with the term. That is, one must have knowledge of what type of word the term is (for example, concrete noun or proper name). These are syntactic markers. One must also have knowledge such as that the term is the name of an animal or mineral or element. These are semantic markers. In general one must have some idea of the type of word the term is; this information is give by semantic and syntactic markers. (2) One must associate the word with a certain stereotype. This stereotype is very similar to what that traditional view called the meaning of the term. It is a list of properties that are associated with the term. These properties are not necessary nor sufficient conditions for being a member of the natural kind referred to by the term. It is not necessary that all or even most of the members of the natural kind possess these properties. They are simply properties associated with the term and the natural kind. (3) One must use the word to refer to a certain natural kind with a certain essential nature.

                It is not necessary for a linguistically competent speaker to be able to decide what individuals are members of a natural kind. An expert who knows certain tests derived from the best theory of the time can decide what things the term refers to if a problem arises. For example, in order to know how to use the term 'aluminum', one need not know how to distinguish it from all other metals. One need only intend to refer to the natural kind that samples of aluminum belong to. (Putnam 1970, pp. 149-52; Putnam 1973, pp. 204-5)

                The main addition to these views provided by "Explanation and Reference" (Putnam 1973) is an explanation of the conditions under which a speaker using a term refers to one set of things rather than another. The reference of a natural kind term is determined by a causal chain between the use of the term and the introducing event. An introducing event is one in which a term is introduced to the speaker through an approximately correct definite description. For example, if one had been around when Benjamin Franklin performed the experiment on electricity, Franklin could have told that person that electricity is a physical quantity that behaves like a liquid, collects in clouds , comes down from the clouds in lightning, and runs along a metal kite string. This would have allowed the person to use the term 'electricity' as well, and all further uses of the term by that person would be causally linked to this introducing event. Even if a person forgot the introducing event, they would still intend to refer to the same thing referred to in the first uses of the term. (Putnam 1973, p. 200) Putnam's theory here diverges significantly from that given by Saul Kripke (1972). The causal chain that determines reference goes back to an introducing event, not to an event that involved an actual member of the natural kind. This is because, in practice, no one will be available to give a definite description of the natural kind unless they are in some way causally connected to an event that involved the natural kind and because the nature of the causal chain between the introducing event and the natural kind does not seem to matter. As long as one can give a description of the natural kind, one can introduce the term. (Putnam 1973, p. 204)

                The reference of the term is determined for all future uses of the term causally linked to that introducing event, although it is not determined by any knowledge that the speaker has. Nor is the reference determined by direct contact with the natural kind by the individual speaker. The knowledge and the contact with the natural kind that determines reference is possessed only by the linguistic community as a whole. Putnam says: "...the use of a natural kind word involves in many cases membership in a 'collective' which has contact with the natural kind, which knows of tests for membership in the natural kind, etc., only as a collective." (1973, p. 205)

                This, then, gives us an idea of what kind of theory to expect in "The Meaning of 'Meaning'". Let us review the theory as we have it so far. The meaning of a natural kind term is not given by a list of properties that are analytically true of all members of the natural kind. Knowing the meaning of a natural kind term involves: (1) knowledge of the semantic markers, syntactic markers, and stereotypes associated with the term; and (2) causal connection with an introducing event (and ultimately to the natural kind itself). The reference of the term is not fixed by the individual, but only by the knowledge and causal connections of the linguistic community as a whole. For the details of this view, and for the arguments supporting it, we must turn to "The Meaning of 'Meaning'".

                That paper (Putnam 1975) begins, again, with Putnam attacking the traditional view of meaning. The traditional theory of meaning rests on two unchallenged assumptions:

                (1) Knowing the meaning of a term is simply begin in a certain psychological state. The traditional view was that the meaning of a term was a bundle of properties. Whether this bundle took the form of a concept in the mind (Locke) or a mind independent sense (Frege) or proposition (Russell), understanding the meaning was still thought to be a psychological state. The type of psychological state that is involved in grasping the meaning of a term is severely limited by the tradition. The tradition, following Descartes, made an assumption about psychological states that Putnam calls methodological solipsism. They assumed that no psychological state presupposes the existence of anything except the subject having the psychological state, not even  the subject's body. Putnam calls such states psychological states in the narrow sense (PNS). (1975, pp.219-220)

                (2) The meaning or intension of the term, i.e., the list of properties that define it, determine the extension of a term, the set of individuals to which the term applies. It was admitted that two terms with different intensions (e.g., 'has a heart' and 'has a liver') could have the same extension, but sameness of intension implied sameness of extension. This is because it is the list of properties that constitute the intension that determines what a term refers to. (1975, p. 217-219.)

                It is a consequence of these two assumption that the psychological state (in the narrow sense) of the speaker determines the extension of the term he uses. Since the PNS of the speaker determines the intension of the term and the intension determines the extension, two speakers in the same PNS cannot be referring to a different set of objects. Putnam argues that this is false, and, because of this, that no theory of meaning can satisfy both of the assumptions. (1975, pp. 219-22)

                Putnam's arguments take the form of a series of more or less unusual examples and some intuitions about these examples. My method in examining these arguments will be to distinguish the different examples, to clarify exactly what set of circumstances the examples are meant to exhibit, and to determine exactly what the intuitions concerning these circumstances are based upon. This will help not only to clarify the position these arguments are meant to support, but also to see how far the arguments support that position. Putnam uses at least five different types of examples to support his thesis that psychological state does not determine extension. The most common scenario Putnam uses as a stage for his examples is an imaginary planet in a far away galaxy called Twin Earth. The planet is exactly like Earth in all respects except these noted in the examples. We are even asked to imagine that we have exact copies on this  planet with exactly similar languages, personal histories, and thoughts. The main difference between Twin Earth and Earth is that on Twin Earth there is no H2O. Instead there is a liquid that is exactly like H2O in all obvious respects, which has a complex chemical formula summarized as XYZ. Wherever we have H2O on Earth, they have XYZ, and they use the term 'water' to refer to XYZ in the same way we use it to refer to H2O.

                The first three examples use this scenario. In the first of these a space ship from Earth visits Twin Earth. At first the visitors suppose that 'water' means that same on Earth and Twin Earth. This will be corrected when it is discovered that 'water' on Twin Earth is XYZ. The visitors then would say that 'water' means something else on Twin Earth, XYZ. In the second example, the exact opposite occurs. A traveller from Twin Earth visits Earth. This visitor makes a supposition and correction exactly parallel to the first case. In the third case the time of the example is changed to 1750, when the molecular structure of water was not known on either planet. An individual on Earth, OscarE, and his counterpart on Twin Earth, OscarTE, would have exactly the same ideas about what they called 'water'. In fact, we could suppose that their psychological states (PNS) are exactly the same. We would still say, however, that the extension of 'water' on Earth is different from the extension of 'water' on Twin Earth. (1975, pp. 223-4) In these examples there are two completely isolated linguistic communities which speak the same language. They use exactly the same syntactic rules. They have exactly the same vocabulary, and the uses and semantic relationships of this vocabulary are remarkably similar on both planets. This, of course, is because the planets are remarkably similar in all other respects, with the one exception being the chemical structure of the material referred to by the term, 'water'. The intuition in these cases is that this term, 'water', would have different extensions in these different linguistic communities. So a speaker from the Earth community would only refer to H2O by 'water', and the speaker from the Twin Earth community would only refer to XYZ by 'water'. Therefore, 'water' has different meanings (in some preanalytical sense) in the two communities. This is so even though the communities are exactly the same when they use 'water'. In the first two cases, experts on each planet had different concepts associated with 'water' in virtue of their knowledge of molecular structure. In 1750, there were no such experts on either planet.

                The view that these examples are meant to support and exhibit is that what makes something a member of a natural kind referred to by some natural kind term, W, is that the individual bears a certain sameness relation (for example, x is the same liquid as y, or x is the sameL as y) to most of the stuff that the linguistic community has on other occasions called W. (Putnam 1975, p.225) Whether or not something bears this relationship to this stuff is to be determined theoretically. Even if all the speakers, including the experts, are unable to determine what bears this relationship, or if they are wrong about what does, the reference of the natural kind term is still the same. It is still that set of objects that actually bears the appropriate sameness relation to most of the stuff we call by that natural kind term.

                Before we consider the implications of this part of the paper we need to consider two other examples that putnam gives to further support his thesis that psychological states do not determine the extension of natural kind terms. These examples differ, in important respects, from the ones we have looked at. The fourth example Putnam uses involves Twin Earth once again, but this time the metals aluminum and molybdenum are involved instead of water. We are to suppose that aluminum and molybdenum are very similar and can only be distinguished by an expert. On Earth, aluminum is much more abundant than molybdenum, and the term 'aluminum'  is used to refer to aluminum while 'molybdenum' refers to molybdenum. The opposite is true on Twin Earth; molybdenum is by far the most common there, and 'aluminum' refers to it on Twin Earth. The intuition in this case is the same. Even though the non-expert speaker on Earth has the same concept associated with 'aluminum' as his counterpart on Twin Earth has associated with it, they refer to different things. Putnam also gives a similar example which does not involve Twin Earth. If all a person knows about beeches and elms is that they are trees, then the concept they have associated with those terms is the same. These terms refer to different things, however, and this difference in extension is not brought about by difference in concept. (1975, pp.225-26)

                In these last two examples each speaker is causally connected to both of the natural kinds involved. This makes them different from the first three examples we considered. This difference is significant because it brings up the question of how reference becomes fixed to a particular natural kind. Presumably, when the reference of 'aluminum' was fixed on Earth there were no experts who could distinguish aluminum from molybdenum. Yet the reference of the term was still fixed only to aluminum and not molybdenum. In the first example. it is easy to see why the visitor to Twin Earth did not conclude that there were two fundamentally different types of water upon discovering that what was called 'water' on Twin Earth was actually XYZ. The use of the term 'water' by the visitor from Earth had no causal link to XYZ. In this example it is more difficult to see why, upon discovering that small amount s of what they had called 'aluminum' had a different atomic structure than  the rest, people did not simply conclude that there were two types of aluminum. The example is not clear about how the reference was originally fixed, so it is not clear if Putnam's intuition that  'aluminum' and 'molybdenum' have different extensions on Earth is based on the fact that experts can distinguish between the two metals now or on the fact that the use of 'aluminum' is causally linked to a group of metal specimens that were mostly aluminum.

                There is also another significant difference between these last two examples and the first ones. In these last examples, although the non-expert speaker cannot identify the difference between the two kinds, there are experts who can identify members of each of the natural kinds involved. Because of this, these examples support the view that there is nothing about the individual's psychological state that determines the extension of a term. It is only the knowledge of the linguistic community as a whole that determines the extension. Putnam calls this a division of linguistic labor. (1975, pp. 227-29) Not everyone who uses natural kind words must know how to distinguish each natural kind from all others.

                Let me summarize what these examples we have considered so far show, if the intuitions Putnam has with respect to them are correct. First, the main task of these examples was to show that psychological state (in the narrow sense) does not determine extension. These examples accomplish this in two ways: (a) In the last two examples, no knowledge that the individual speaker has determines the extension of the terms; only the linguistic community as a whole has this knowledge. Therefore, the PNS of the individual does not determine the extension of terms. This is a result of Putnam's thesis that reference involves a division of linguistic labor. (b) The first three examples show that extension is not determined by psychological state in a more radical way. They show that extension is not determined by the psychological states of the linguistic community as a whole. Second, the examples suggest that what determines the extension of a natural kind term is the essence of the natural kind to which the term is linked. It is not necessary for the individual to be able to decide what things belong to a natural kind in order to be a competent speaker. There are experts available to decide what things are in the extension of a term. This supports the view that an individual belongs to a natural kind if it bears the appropriate sameness relationship to those objects we usually use the term to refer to. Third, they suggest that the particular natural kind to which a term refers is determined by a causal connection between the term and the natural kind. This supports the theory considered above that a causal connection between the term and an introducing event fixes the reference.




Putnam, Hilary. 1970. "Is Semantics Possible?." In Mind, Language and Reality, Cambridge: Cambridge University Press, 1975, pp. 139-152.


  1973. "Explanation and Reference." In Mind, Language and Reality, Cambridge: Cambridge University Press, 1975, pp. 196-214.


  1975. "The Meaning of 'Meaning'," In Mind, Language and Reality, Cambridge: Cambridge University Press, 1975, pp. 215-271.