I've finally gotten around to looking at "Truth and Predication", which is Davidson's first and posthumous book. Up until now, the only place I've seen the book so much as mentioned was in its NDPR review from when it first came out; I've wondered why it seemed neglected.
It turns out that the first half of the book is a reprinting of Davidson's Dewey Lectures, previously published as "The Structure and Content of Truth". I had thought it was odd that these lectures hadn't been included in any of the Davidson essay collections; mystery solved. This also explains why the book hasn't been mentioned much: Half of it is old material.
Davidson mentions in the Introduction that he's left the lectures basically untouched, since they've already been widely cited & commented upon; he says that the few changes he's made have mostly been in footnotes. Apart from one marginal note at the end of the third lecture, I only found a single revision of possible substance.
Here's the footnote as it appears in chapter three of "Truth and Predication", p.51: "The step from observed assents to inferred attitude [sic] of holding true is not, I think, in Quine."
And here it is in "The Structure and Content of Truth" p. 318: "The step from observed assents to the inferred attitude of holding true is not, I think, explicit in Quine." (my emnphasis)
Now, given that the sentence has lost an article in the transition from article to book, it's possible that this is just a printing error. If so, it's an amusing one.
(Well, I was amused.)
The marginal note at the end of the third chapter (which equals the last of the Dewey Lectures) is obscure; Davidson was noting some things he wanted to incorporate "in chapter 2 or 3". But there is this rather nice bit in it: "I want to make clear that my 'solution' isn't a basic one. It is an alternative to deflationary, epistemic or correspondence theories not in proposing a better definition (or short summary) but in suggesting a different approach which relates the concept of truth to other concepts."
I haven't gotten around to reading the "Predication" parts of the book yet, though the introduction makes clear that Davidson thinks that something Tarskian will do the job (and nothing else has -- most of this part of the book is historical/critical).
The introduction also features Davidson excusing himself from addressing the semantic paradoxes. I'd wondered what he had to say about those; it did seem a little odd that he'd written so much about truth without addressing "this sentence is not true". I'll just reproduce the passage, because I am too tired to summarize and I have to get up early in the morning to move things into a truck:
I have been chided more than once for leaving out the semantic paradoxes. The honest reason is that I have nothing new to say; I like the proposals of Burge and Parsons. How can I say the concept of truth is so clear? Well, relatively clear. The paradoxes don't intrude in our ordinary talk. Why not? They arise when we try to assign truth values to sentences containing the concept of truth. But sentences are already a long way from most ordinary speech. We don't utter sentences, but rather tokens of sentences. Since communication depends on what we make of the tokens of others, and communication often succeeds, we can normally assume that others mean what we would mean if we uttered those sentences. This is something we can and do check up on, consciously or not, all the time. But it remains the case that we succeed only to a degree (there are many dimensions). Truth, whether of sentences or of utterances, is relative to a language, and we never know exactly what the language is.
It is not my view that therefore the concept of truth is ambiguous. No more, anyway, than in the case of any word. Our words are clear enough in the circumstances in which they have been used. When we test the limits, we are typically not asking "what does it mean?" but "how shall we use it now that these difficulties have come up?"
As for sentences without a truth value, and names without a reference: again, this is a topic on which I do not feel I have any serious and original thoughts. We know the semantic role of names that do refer; it's one of the first things we learned. But this is of no help in deciding whether sentences containing proper names have a truth value. Our intuitions, based on our knowledge of their role when they do refer, prompt one (me) to hold a sentence like 'Zeus does not exist' as true if there is no one who fulfills certain usually adequate properties, and false if someone does. But I intuitively treat the sentences in Homer that recount some of Zeus's sexual misbehaviour as neither true nor false. But of course the context is all. I do not mean it is pointless to consider seriously the semantic role of proper names. Just as this book illustrates two different routes into the simplest sentential structures, starting with reflections on the role of proper names might end up doing the same thing. "The Problem of Proper Names" might then have taken the place of "The Problem of Predication." [marginal note here: "Certainly, for Quine; maybe for Russell."]