Although Soames’ paper came 40 years after Tarski’s (and speaks to the views of a couple of people we will read later), we are reading it at this point because the issue Soames addresses is a new version of a criticism of Tarski that was made from the beginning: although Tarski gives a recipe for defining truth in a broad class of languages, any definition along his lines is a definition for only one language, and the concept of truth may seem to properly concern languages generally.
We will spend two classes on this paper. In the part we will discuss first, §§I-II, Soames sketches a motivation for Tarski’s theory and one well-developed criticism of it. In the rest of the paper, §§III-V, he looks more briefly at further criticisms and describes the way he thinks Tarski’s theory should be understood. Although the idea of “physicalism” will come up a lot, I won’t refer to it in the notes below because it really functions only as an example of an attempt to account for facts about meaning; a “mentalistic” approach that attempted to account for meaning in terms of Cartesian thoughts and feelings would lead to analogous issues.
• The key idea to notice in §I is the association between Tarkski’s theory and “redundancy” theories or “deflationism” (think of Ramsey) on the one hand and the idea of “semantic ascent” on the other.
(There is one bit of logical notation in some of the examples: p → q amounts to “if p then q,” so example 3a says “if snow is white, then if grass is blue then snow is white” and might be restated as “if snow is white, then snow is white if grass is blue.” Soames is probably choosing this odd sort of sentence as a random logical truth. It appears as the first axiom in many axiomatic formulations of the logic of conditionals.)
• Section II then looks at a well-known criticism of Tarski’s definition due to Hartry Field. What is probably the clearest statement of Field’s criticism comes at the bottom of p. 419 just before Soames begins his criticism of Field. (In thinking about Soames’ criticism of Field, notice a similarity between the ideas of Quine that Soames mentions on p. 421 and what Ramsey tries to do with the ideas of belief and disbelief in the latter part of his paper.)
If the earlier statement of Field’s criticism using examples (9) and (10) is puzzling, note that Tarski defines the semantic notions appearing in (9) in such a way that (9 a-c) are not informative facts concerning reference, etc., but instead are built into the definition of these semantic notions. It is then no surprise that restating (9 a-c) using the definitions yields the tautologous (10 a-c). The point of displaying the examples (10) is just to show how little the examples (9) really say if the semantic notions are taken to be defined in Tarski’s way.
• Soames considers further analogous criticisms of Tarski more briefly in §III. Here the odd examples (17T and 18T) are different from those in (10) but the problem they point to is similar: because Tarski’s definition only applies to one language as it actually is, it does not provide a concept usable in the sort of “what-if” reasoning exhibited in (17) and (18).
• In the last two sections of the paper, Soames suggests a way of understanding the relation between Tarski’s definition and broader questions about meaning. At the heart of this is the idea of “truth in L for variable L” that he sketches early in §IV. (The qualification ‘for variable L’ indicates that this is not a predicate “true in L” for a specific language but a general concept of “true in” that applies to a range of languages.) Soames doesn’t really depart from Tarski here because this idea matches the way that Tarski’s conception of truth was applied, by him as much as anyone, in the study of models of mathematical theories. The philosophical upshot of this way of looking at Tarski’s definition is summarized in the first paragraph of §V and in the latter part of the last paragraph.