An Epistemic Problem for Intentional Semantics

By Travis McIntyre

Abstract: This paper concerns the concept of reference within the field of semantics. W. V. Quine argues in his Word and Object that the relation between words and the objects they refer to is metaphysically indeterminate; there are no facts in the world which can determine what objects words refer to. This paper refutes this thesis by expanding the available facts for establishing reference from behavioral facts (stimulus meaning) to include mental facts which include peoples‟ intentions (intentional semantics). I go on to point out how this new set of facts does not entirely escape Quine‟s indeterminacy thesis. There remains an epistemic problem of knowing what objects given words refer to. This epistemic problem seems to fall under the general skeptical problem of under determination of theory by data. But the paper shows that the epistemic problem is a more significant threat than under determination of theory by data.

Semantics is a crucial aspect of language and much has been written on it since the start f the field. One of the most fundamental concepts within semantics is the reference relation. It is the relation between words in a language and the objects they refer to. Intuitively, this relation is and must be determinate in that there are facts in the world which determine what object a given word refers to. If this relation is in determinate, language would be of little use because what is being referred to would be unknowable. But there are several philosophers who have questioned the determinate nature of language and concluded that language cannot have a determinate reference from words to objects in the world.1

.

The second chapter of W. V. Quine’s Word and Object is devoted to showing that the reference of words to the world is indeterminate. Further, Quine takes the meaning of words to be metaphysically indeterminate: there are no facts in the world which can give words a definite meaning. This paper explains why the metaphysical claim about the nature of language is fallacious. But while the metaphysical claim is fallacious, there remains a genuine epistemic problem about the inability for someone to know what someone else is referring to when using a particular word. This problem reduces to under determination of theory by data, so in order for the epistemic problem to be significant there must be a difference between the under determination of theory by data found in the sciences (e.g., physics, biology, etc) and that of semantics. This paper shows that a significant difference between the under determination of theory by data found in the natural sciences and that of semantics exists. Therefore it shows that Quine‟s case for metaphysical indeterminacy entails a genuine epistemic problem for the relation of reference from words to objects in the world.

.
Quine begins his argument by outlining a task he calls radical translation. Radical translation is the process of creating a translation manual for a completely unknown language. By a translation manual Quine is thinking of a comprehensive list of the words in the unknown language mapped onto the words of the known language. Section 15 begins with a summary of the method employed by the field linguist in undertaking radical translation: “We have had our linguist observing native utterances and their circumstances passively, to begin with, and then selectively querying native sentences for assent and dissent under varying circumstances.”2 This method yields the stimulus meaning of the queried words, which Quine takes to be the only objective notion of meaning. Stimulus meaning is the class of all stimuli which would prompt someone to assent to or dissent from a sentence.3 By virtue of establishing the stimulus meaning of certain sentences the reference of a particular class of sentences can be established. This class is composed of those sentences which command assent or dissent when queried after an appropriate stimulation, and are unaffected by collateral information, where collateral information is taken to mean additional information which is not obtained during the prompting of assent or dissent. This type of sentences is called observation sentences, and for these sentences Quine’s method is able to establish their meaning.
.
Further, the field linguist needs to get from the meaning of sentences to the meaning of individual words. And if the meaning of individual words can be established, then the occurrence of these words in other types of sentences will have meaning and thus give these other sentences meaning. Quine’s thesis of indeterminacy of meaning takes effect when one extends the meaning established for observation sentences to the meaning of the words within these sentences.
In order to get from the meaning of sentences to that of words, the field linguist must make what Quine terms analytic hypotheses. Analytic hypotheses consist of breaking down sentences into words and giving the words tentative translations.4 The step of making analytic hypotheses is where Quine’s skeptical claim about language arises. This is because there are multiple translations of each word which are consistent with the stimulus meaning of the observation sentences they occur in but which are incompatible with each other. To illustrate this, Quine uses the example of “Gavagai” and “gavagai” where “Gavagai” refers to the sentence and “gavagai” refers to the word. The sentence “Gavagai‟”has the stimulus meaning of assenting to stimulations in which rabbits run by and dissenting to stimulations which do not include rabbits. The commonsense translation of “gavagai” is “rabbit” but Quine points out several alternative translations: enduring rabbit, temporal slice of a rabbit, undetached rabbit part, mereological fusion of all rabbits or the property of rabbithood.5 All of these translations are consistent with the stimulus meaning of “Gavagai” but are inconsistent with each other. And if a word doesn’t have a single translation or, if there are multiple translations, they aren’t inconsistent with each other, then the word’s meaning will be indeterminate. And since all words can be given different yet incompatible translations (such as was given for “gavagai”), all words suffer from indeterminacy of meaning.
.
It is important to notice the type of claim Quine is making. It seems that it could be an epistemological claim about the field linguist‟s inability to determine the meaning of the native words. But a much stronger claim is actually being made: a metaphysical claim about the nature of language itself. Quine asserts that language itself is such that words have no determinate meaning.6 This is a very strong claim in that it extends not only to all words but all natural languages. Quine is able to extend the indeterminacy of the words in the case of a completely unknown language, such as was encountered during radical translation, to all languages because he is a strict adherent of behaviorism. This belief in behaviorism explains why he takes stimulus meaning to be the only objective notion of meaning; behavioral facts are the only facts which can give meaning to words. Language is a human convention and therefore the meaning of words must be based on human conventions. Since these conventions arise from the mental capacities of people, we must look to the mental capacities of people to account for the meaning of words. But according to behaviorism, the mental capacities of people can only be characterized by behavioral data such as stimulation and reaction. Therefore, if the meaning of words is determinate, it must be sought through behavioral facts. But behavioral facts cannot establish a definite referent for a given word. Even if every behavioral fact is established, they still cannot distinguish between different, incompatible translations for a given word.7 Thus Quine’s argument of indeterminacy extends from cases of radical translation to all languages. By this line of reasoning and the presupposition of behaviorism, Quine‟s metaphysical thesis about language is shown to be true.
.
This metaphysical claim is quite easy to reject. It relies on behaviorism which is itself a controversial position. So by rejecting behaviorism, the metaphysical claim about language is also rejected. To do this, behaviorism has to be shown to be fallacious and another source of facts which establish meaning must be given. A strong argument against behaviorism is to point out the many counterexamples in which one’s behavior is contrary to their mental states. For instance, people have a tendency to hide their feelings; whether it is pretending to like someone you actually despise or telling people you are happy when you are not, the behavioral data does not match the mental state of a given person. The existence of counter examples is sufficient to reject behaviorism. If behaviorism is rejected, then mental activity would be a legitimate source of data to be to establish meaning. So the intentions of someone using particular words can constitute legitimate data to establish the referent of those words. So what one intends to refer to by the words they use constitutes the meaning of those words. A problem with this source of meaning is that it doesn’t take into account that different people can intend to mean different things. Therefore it will help to relativize intentions to a language shared by a given community whose members intend the same thing by the words they use. So, what a given community intends to mean by a word it uses constitutes the meaning of the given word.
.
This account of how the reference relation is established gets around Quine’s problem. By using the word “gavagai”, the native intends the word to have a certain meaning. Thus, regardless of what the native intends to mean by “gavagai”, it will be determinate. Words will have a determinate meaning if intentions are accepted as the facts which establish the meaning of words. I will term this intentional semantics and take it to be an alternative to stimulus meaning. Its end (like stimulus meaning) is to capture the intuitive notion of meaning (referential semantics). Therefore Quine‟s metaphysical claim about the nature of language is properly rejected. It has been shown that there are facts in the world which secure the reference of words to objects in the world.
.
Even after we have successfully refuted the metaphysical claim made by Quine, a problem remains. What the native intends to mean by a given word is unknowable to the field linguist. In so far as a person is able to know what another person”s mental state is like (which includes their intentions), it must be by virtue of communicating through language what their mental states are like. This is because mental states are not directly observable. Similarly, knowing what ones pet dog is thinking is difficult because they do not share a common language with their dog. Thus there is no way to determine the intentions of the native since the language employed by the native is completely unknown to the field linguist. Since meaning is derived from the intentions of the speaker, if the field linguist doesn”t know the intentions of the speaker, he can‟t know the meaning of the words used by the speaker.
.
This line of reasoning supports the use of behavioral facts and therefore behaviorism. If we cannot know the mental activity of the natives through language, our best method would be to get at their mental states through behavioral data. One is able to tell that their dog is hungry when it spends all day sniffing around its food bowl and licks old food bits off the kitchen floor. Thus we use behavioral data to tell what our pets are thinking. Similarly, in order for the field linguist to have knowledge of the native’s intentions, the linguist must resort to behavioral data. And this ultimately leads back to Quine‟s thesis of indeterminacy of meaning. The linguist has multiple, incompatible interpretations of the native’s intentions. Based on behavioral data the linguist cannot determine if the native is intending an enduring rabbit, a temporal slice of a rabbit or rabbit hood as the meaning of “gavagai”. Any of these interpretations are consistent with the behavioral data, but all of them are incompatible with each other. Although it may seem as though we went around in a circle and didn‟t achieve anything by identifying meaning with what one intends to mean, we have changed the nature of the argument: we are now faced with an epistemological problem rather than a metaphysical one.
.
This epistemic problem follows a general form of skeptical argument: under determination of theory by data. This form of argument is applicable to any theory building discipline, and is often applied to the natural sciences. The argument states that for any set of data, there are multiple, incompatible theories which account for the data. So, for instance, in physics there is the established theory of gravity: a force which pulls everything toward the earth at a constant rate. This theory was established to account for the data of objects falling to the earth at a constant rate. But there are other theories which also account for the data: e.g., there are invisible people who jump up and down on objects at a constant rate such that everything is covered with them and everything falls at a constant rate. This is an absurd theory but it does account for the data of the objects falling at a constant rate towards the earth. As under determination of theory by data is presented in this case for physics (as well as most of its other instances within other fields of science), it does not seem to be a very serious epistemic problem. It just seems to be a skeptical hypothesis, such as the brain in the vat scenario, which, while it cannot be ruled out, it is not taken seriously either. So it is of the utmost importance to see if the epistemic problem based on Quine‟s argument for the indeterminacy of meaning reduces to the weak skeptical hypothesis of under determination found in the sciences. If it does, then it is not a serious epistemic problem. But if it can be shown to be essentially different from the weak skeptical hypothesis, then it can be maintained as a serious epistemic problem within the philosophy of language.
.
In analyzing skepticism of the reference relation, Chomsky formulates an alternative methodology of theory building.8 Given a phenomena and related data, a theory about that phenomenon is given. Then, in accordance with future data, the theory is accepted, modified or rejected. The theory itself, when supported by data, justifies belief in unobserved facts (these can be such facts as gravity in physics or intentions in intentional semantics). Furthermore, since different yet incompatible theories can be made for a given set of data, a hierarchy of better and worse theories can be given. This normative action can be based on additional data, simplicity, plausibility, etc.9 So, in the case of the incompatible theories of falling bodies which were given above, one should be better than the other. With respect to simplicity and plausibility, the theory of gravity is much better: not only would having lots of invisible things be bad for parsimony, but an account of the invisible people would be required. What do they eat? When do they sleep if they are always jumping up and down? Do they have souls? Although these are ridiculous questions to ask, some answer should be given if this is to be a plausible theory. Thus it seems that gravity is a much simpler and more plausible theory. So the different theories proposed by the weak skeptical hypotheses of under determination of theory by data conform to being better and worse theories. In opposition to this, the alternative interpretations of the native‟s intentions are not hierarchical.
.
Before explaining why the alternative interpretations of the native’s intentions are not hierarchical, we should look at the maximum benefit of Chomsky’s method of theory building. If it was the case that one interpretation was better than the others, then the belief in that being the intended meaning of the word would be justified even though the native’s intentions are not directly observed. This is because Chomsky says that a good theory justifies belief in unobserved facts. But in order for this to be the case, there must be a best interpretation of the native’s intentions. This is not the case because there is no criteria by which to judge one interpretation as better than another. Because the different interpretations are based on different metaphysics, it is hard to compare their relative simplicity and plausibility. Further, since the epistemic problem arises even when all behavioral data is assumed to be available to the linguist, additional data cannot rule out an interpretation; not only do all the interpretations account for the behavioral data equally well, but there is no increasing sum of data since it is assumed to be had by the linguist in its entirety. Thus it seems that a hierarchy of the possible interpretations of the native‟s intentions cannot be given. This is the essential difference between the under determination found in the sciences and that of semantics: the alternative theories of a given phenomena in the sciences are hierarchical where as the alternative interpretations of the natives intentions are not. Thus the epistemic problem discussed in this paper is a genuine problem for knowing what someone is referring to with a given word being used.
.
One aspect which needs elaboration is the role of metaphysics in the alternative interpretations which can be made by the field linguist and the possibility of using metaphysical arguments to decide which interpretations are better (in Chomsky’s hierarchical sense). There is a strong reliance on metaphysics in the alternative interpretation of the intentions of the natives. For example, the alternatives of an enduring rabbit and temporal slice of a rabbit are based on the metaphysical theses of endurance and perdurance. When it was wrote that the different interpretations are based on different metaphysics and it is therefore hard to compare their relative simplicity and plausibility, this statement was incorrect. There are endurantists and purdurantists because metaphysicians give arguments as to which is a better theory. Thus one could use these arguments to decide which interpretation of the natives intentions (whether the native intends enduring rabbit or temporal rabbit slice by “gavagai”) is superior. And there would then be a better and worse interpretation of the native‟s intentions. But using this methodology would presuppose an unwarranted knowledge of metaphysics upon the natives. What a native intends a given word to mean would ultimately be based on their cultural upbringing. If, for whatever reason, the native‟s culture adhered to the metaphysics of purdurantism, then “gavagai” would refer to a temporal slice of a rabbit rather than an enduring rabbit. But this would not be a view held because of the arguments found in modern metaphysics. So to infer native intentions from metaphysical arguments would be off base.
The arguments presented make a case for the under determination of the relation between words and the objects they refer to within radical translation. Rather than being a metaphysical problem as was found in Quine’s original argument, I have made it a purely epistemic problem. But this thesis can be expanded from radical translation to all languages in the same manner that Quine was able to expand his metaphysical thesis. Thus the epistemic problem contains the full force of Quine‟s original argument except it does not rely on the controversial position of behaviorism.
.

Notes

.
1 Two well known papers on this topic are: W.V. Quine, Word and Object (MIT Press, 1960) and Hilary Putnam, Reason, Truth and History (Cambridge University Press, 1981). This essay concerns the first of these papers.
2 Quine; pg. 68
3 Quine; pg. 32-33
4 Quine; pg. 68
5 Quine; pg. 51-52
6 Quine; pg. 73
7 Quine; pg. 72
8 Noam Chomsky, Knowledge of Language: Its nature, Origin and Use (Praeger, 1986), pg. 249-250
9 Chomsky; pg. 250
.
Travis McIntyre (’10) is a Philosophy major at University of California Davis.

Art courtesy of maxspider.

Leave a Reply

Your email address will not be published. Required fields are marked *