Remembering Donald Davidson: His 1967 Undergraduate Philosophy of Language Course
During his first year at Princeton, Fall semester, Donald Davidson gave an undergraduate philosophy of language class, attended by several graduate students. We graduate students were not supposed to ask questions, but we could listen and takes notes.
For me, Donald was first and foremost, my teacher. So, rather than reminisce about the various interactions with Donald I had over the subsequent decades, I will instead sketch some of what went on in that undergraduate class. The following sketch seems to me to show more about what was great about him, as a philosopher, than any array of my anecdotes would. In this 1967 class you see a great mind sharing original ideas and perspectives which often only much later find their way into print. In the following, quoted sentences are intended as direct quotes of what I wrote down as Donald’s words.
II Fall 1967
Looking at the notes I took forty-four years ago, I see that the opening class (September 19) was a discussion of what language is and what philosophy of language is. In the course of a discussion of why philosophy of language is important, and what language is not, Donald presents the basic conception of “The Method of Truth in Metaphysics” (1977). In that same class, he gives a central argument of “On the Very Idea of a Conceptual Scheme”(1974) that people arguing that there are incommensurable “conceptual schemes” cannot be right, given that they are explaining those schemes. The next class, on September 21st, was devoted to the semantic paradoxes. This topic was continued on the 26th, with much of the discussion that eventually appeared in “Quotation” (1979).
On September 28, Donald presented a proof using Godel-numbering that the semantic paradoxes do not obviously depend on self-reference. Later that class, an argument that the concept of truth is indefinable shows up which eventually appears in “The Folly of Defining Truth”(1996).
On October 3, Donald discusses the concept of language, exploring Wittgenstein’s “block-slab” language, maps, Bennett’s discussion of the language of the bees, musical scores, and demonstratives. October 5 is devoted to J. L. Austin’s “How to Do Things with Words,” appreciating this masterpiece, but noting a conflation of action-type-predicates with properly semantic predicates—what a person is doing with a sentence versus what the sentence’s truth-conditions are. Some of this material shows up in articles that appear in the late seventies, including “What Metaphors Mean”(1978), and “Moods and Performances”(1979). The October 5 class also contains with the central insight that force cannot be encoded in words, that there cannot really be conventional force-markers, since any words can be spoken without the conventional force, for instance on the stage or as an example. While he did not cite Tractatus 4.442, it is clear that Donald thought Wittgenstein’s insight had been lost on subsequent thinkers about language.
The October 10 class developed the Fregean idea that the sentence is the fundamental unit of meaning, so that words are abstracted from sentences, as developed and extended by Quine. The general methodology of getting the literal meanings (truth-conditions) of sentences and then constructing what a person has in mind by using a sentence was explained, as well as the hopelessness of trying to have a theory that went in the other direction, from a person meaning something by a speech act to what the words mean. Roughly, that version of “meaning is use” cannot be the route to an adequate semantics. The important distinction between denying that p and asserting the negation of p (act-characterization versus sentence-characterization), and so the difference between contradicting oneself and asserting a contradiction, ended that class.
The classes on October 12 and October 17 returned to the discussion of speech acts and truth-conditions, and examined the differences between representations and sentences. “The sentence/world relation is not like the picture/world relation at all.” The discussion of facts argued that if facts were complex entities in the world, it shouldn’t matter how those entities were characterized. This train of thinking continued in the October 19 class, which was devoted to arguments against the idea that there are any entities corresponding to true sentences. Much of the class was taken up with versions of what has come to be called the Slingshot.
The difficulty with getting the point of the Slingshot across is that a closed sentence has to be a component of a singular term. The version Donald presented in class is one I have never seen elsewhere. The context is that not everyone with tickets will show up for a game unless it is not sunny, whereas all and only those who hold tickets will show up if it is sunny. “1) Suppose sentences have a semantic connection with facts. (referring, naming). 2) `It’s sunny’ pictures the fact that it’s sunny. 3) Logically equivalent sentences picture the same fact. 4) If it’s sunny, then the ticket-holders = the game-goers. If it’s not sunny, then there are no game-goers. So, the class of game-goers = the class of ticket-holders and it’s sunny. So, `The class of ticket holders and it’s sunny = the class of ticket holders’ is true if it’s sunny and false if it’s not sunny, since then the class of ticket-holders and it’s sunny is the null class….” 5) `It’s sunny’ is logically equivalent to `The class of ticket holders and it’s sunny = the class of ticket holders.’ “6) If sentences picture facts, [they] will picture the same fact if you substitute for a singular term a term that names the same thing. Whatever fact makes `Venus is a planet’ true makes `The Morning Star is a planet true. By logical equivalence, the class of ticket holders and it is sunny = the class of ticket holders. 7) “Those who hold tickets and 2+2=4 are those who hold tickets” ought to name the same fact, provided it’s sunny…..8) 2+2=4 pictures the fact that it is sunny.” What is different in Donald’s presentation is that the insertion of a sentence into a singular term is intuitive, rather than sounding like a logician’s trick.
The class on October 19th concluded with an exposition of Davidson’s notion of truth, via a discussion of Austin and Strawson on truth. Austin and Strawson had construed the issue of “correspondence” as the question of how to understand a sentence like “The statement that the cat is on the mat corresponds to the fact that the cat is on the mat.” Austin takes “corresponds to” and so “is true” as a relational predicate linking the statement with the world. Strawson treats “is true” as a pure performative, so that “It is true that the cat is on the mat” is not relational at all. Donald agrees with Strawson that “is true” is not a relational predicate, but agrees with Austin that it is a semantical predicate. “Truth is a one-place predicate.” Donald continues by pointing out that part of the confusion is between speech-act predicates and semantical predicates.
The October 24th class draws consequences of the rejection of facts for logical form, some of which show up in “Truth and Meaning”(1967) and later essays. The primary difficulty that the Slingshot poses is how to understand apparent sentence connectives which ought to be referentially transparent contexts, but where the component sentences must contribute something other than their truth values. Donald’s example was “Jack fell down before Jack broke his crown.” He points out that sentences are not names of events, even though “Jack fell down” might seem to be so. Rather than go on to give event analyses, though, the rest of the class, and the next several classes, dealt with logical form, the requirement of learnability, what is part of form and what is not, the idea that a semantics need not be in terms of referents, and other equipment for understanding what “Truth and Meaning”(1967) is about.
The class on October 26 separated the systematic study of meaning from the analysis of speech acts, using his accounts of intention and multiple predicates for the same event, and suggesting that the project of trying to get an account of meaning by working backwards from sounds and actions performed, without going through a theory of truth-conditions, would be a very ambitious behaviorist project. Could one even give an account of intensions on the basis of intentions? The October 31 class continued the discussion of linguistic meaning and its separation from questions about the uses to which words can be put. Linguistic meaning is public and learnable in advance.
The classes of November 2 and November 7 dealt with the contrasts with and connections between Davidson’s semantic project and Chomsky’s syntactic project, first making clear what syntax was supposed to do, and then arguing that much of “deep structure” was really about logical form. In particular, the pair “I expected John to leave” and “I persuaded John to leave” were argued to differ primarily in logical form, and that essentially semantic considerations, rather than strictly syntactic considerations, were driving syntax. This was illustrated by considering what happens under a passive transformation. “The pair `I persuaded a specialist to examine John’ / `I persuaded John to be examined by a specialist’ have different cognitive significances, unlike the pair `I expected a specialist to examine John’ / `I expected John to be examined by a specialist.’” Donald emphasized that getting logical form right was a precondition of getting a correct syntax for a language.
The rest of the class of November 7 continuing into the November 9 class began the topic of how to begin a recursive account of the meanings of the sentences of a language. Much of the material in “Truth and Meaning” is in my notes. Donald discussed the great advantages of Frege’s treatment of predicates as functions, so that one avoids the problem of “gluing together Theaetetus and flying” that treating predicates as entities involves. Just as we don’t need “the father of” to name an entity, but just need mapping from entities to their fathers, so a semantics need not assign entities to predicates—the clause in the truth-definition will do the work.
On November 14 Donald presents an argument which I do not remember seeing in his published work: If we treat predicates as we treat functions, “a complex expression channels reference down…Each time a functional expression is applied, the scope of what is referred to, what the sentence is `about’, is reduced.” “Meaning” as an entity, is cut down and simplified. So, calculating the reference of the sentence, “Jack fell down and broke his crown” you have the “and” function applied to the pair of the result of applying the “fell down” function to Jack and the result of applying the “broke” function to the pair the “crown of” function applied to Jack and Jack. Each application reduces the things talked about, eventually getting to a single thing. That is, And(Fj,B(Cj,j)) comes down to And (T,T) and that comes down to T. So, where we started out with a number of entities, Jack and his crown, the function-argument idea ends up with a unitary object, a truth-value.
The same thing should happen if senses were taken to be meanings that were entities. The sense of a complex expression would be a function of the senses of its components. The functions likewise would reduce the number of entities referred to, eventually paring the entities down to one, the meaning of the whole complex. So, the senses of propositions would be unitary entities. So, if you had meanings you could not have a theory of how they are got. “Meanings are single entities--they don’t reproduce their genesis and so cannot be the basis for a recursive theory.” If a theory is out to produce entities as meanings of sentences, it is hopeless—“to be told that the sense of `fj’ is `the sense of fj’ isn’t to get anywhere [because you] don’t get structure.” Donald’s discussion then moves on to how a truth-definition does gives structure: “Shows how the truth-conditions of sentences depend on structure.”
The November 16th class discusses Convention T and how not to understand it. “Convention T is a non-constructive characterization of the extension of `Tr’ in any language. It doesn’t even hint how to give a definition.” “A recursive characterization doesn’t show how to get rid of the predicate—so we determine the extension but leave in `is true’. `Is true’ is a primitive notion even though the extension is fixed.”
The class on November 28th, apparently after Thanksgiving break, continued the discussion of applying truth-definitional structure to natural language, arguing that logical grammar cannot be a tree structure. He gave a picture of what a completed theory would be: “Isn’t my theory trivial? No, because it answers what words do in particular places. It gives truth-conditions for an infinite number of sentences by a recursive account and gets real structure, and so it gets real grammar. You have to solve problems of philosophy to get it done. [Logical form] must be the core of an adequate linguistic theory. It answers the question of the relation of language to the world. It relates an arbitrary sentence to the world.”
The class of November 30th dealt with the details of truth-definitions, and ended with a discussion of “Conditions imposed by a truth-definition: A theory can’t treat language as if there were an infinite number of primitives.” “There are lots of places in English where you seem to need an infinite number of primitives. In every part of philosophy, ethics, philosophy of science, etc., you seem to need an infinite number of primitives.”
The December 5 class illustrated how problems in the theory of meaning correspond to philosophical problems. Donald discusses causal statements, ethics (thereby giving me the topic to my dissertation a couple of years later), aesthetics, and epistemology.
That class next dealt with moods. “Imperatives have truth-values, but don’t bring that sort of thing into question. The hearer makes the imperative true or false.” Donald then emphasizes the difference between a theory of meaning and a theory of use. The class ended with a segue into the account of demonstratives and radical translation. “Demonstratives get you into language in radical translation, so will say sentences are true as uttered by a person at a time.” The truth-predicate will be true of a triple of a person, a sentence, and a time. “Drop the language place.” The thesis of “A Nice Derangement of Epitaphs”(1986) is stated, briefly.
The class on December 7 covered sentences about propositional attitudes and intended to get to radical translation, which Donald did not get to until that December 12 class. Donald set up the problem of propositional attitudes with the most straightforward case, “says that.” “In indirect discourse we compare two different sayings.” He discussed Carnap’s account, Church’s objection, Sheffler’s account, and Quine’s account. He introduced the notion of semantic innocence: “In `Scott said that Venus is an inferior planet’ these words `Venus is an inferior planet’ couldn’t be doing anything unusual.” The class ended with Donald presenting the theory of “On Saying That.”
What appears to be the last class of the semester, December 12, continues the discussion of belief sentences, citing Mate’s argument that no conditions on synonymy will suffice to guarantee that substitution in every context preserves truth. Via his theory of same-saying, Donald explains the peculiarity of performatives.
The topic shifts to radical translation. The topic naturally follows, given the centrality of assertions, and the fact that “on the whole, people can’t be wrong about what they believe, intend or mean. We must take these as true if we think a person is speaking honestly.” Quine’s use of such occasion sentences leads to Donald final discussion of the course, a sketch of something close to part of “On the Very Idea of a Conceptual Scheme”(1974). A nice series of diagrams pictures Quine’s view of language as divided into sentences tied to experience directly and others with conceptions Quine argued against in “Two Dogmas” that have an intermediate level of sentences defined in terms of those immediate observation sentences. He then characterizes his own picture as dropping “the third dogma, that there are sentences free of the network which are directly tied to experience.”
Donald’s diagram of his own view shows no divisions at all. The last sentences of the course make two points about observation sentences and occasion sentences: 1) An occasion sentence like “There’s a rabbit” depends on views and beliefs about rabbits. 2) “Having the same meaning for `rabbit’ for different people doesn’t depend on sameness of stimulation. [Consider] a black box speaking English—its sensory apparatus might be totally different.” All that matters is agreement on occasion sentences. So there is no need for a common observational or stimulational ground.
This was an amazing course from which I learned more philosophy than any other course I have ever taken. It was a great privilege to be there.