Hermeneutic Practice and Theories of Meaning1
1. Two traditions and a common problem
Hermeneutic theories of the practice of textual interpretation and philosophical theories of meaning both address the topic of linguistic understanding. Their motivating problems and corresponding explanatory strategies are quite different, and in some ways complementary, however. Theories of meaning of the sort epitomized by the work of Donald Davidson, Michael Dummett, and David Lewis take as one of their primary explanatory targets the fact that linguistic understanding is in one way surprisingly easy. By contrast, philosophical hermeneutics as developed by Hans-Georg Gadamer takes its point of departure from the fact that some sorts of linguistic understanding are surprisingly hard.
One can begin thinking about the first tradition by recalling the fundamental empirical observation by which Chomsky set the agenda for modern linguistics: almost every sentence uttered by an adult native speaker is novel, not just in the sense that that speaker has never uttered or heard that exact sequence of words before, but in the much stronger sense that no-one has ever uttered or heard it before. Yet it is not just possible for us to understand all these novel sentences, it is generally pretty easy: we all do it all the time. How? There is basically only one idea out there about this: semantic compositionality. Unfamiliar sentences are syntactically built up out of a finite stock of familiar words, according to structures that are finitely generable in familiar ways. Some corresponding fact must hold on the side of semantics. One among many ways to work out this idea is based on Tarski’s recursive definition of a truth predicate for formal languages. Davidson saw, first, that Tarski’s theory could be understood as theory of meaning for the logical connectives associated with recursion clauses in the definition, and second, that this model could be extended to provide a recursive theory of the contribution to the truth conditions of novel sentences that is made by the occurrence in them of arbitrary non-logical sub-sentential vocabulary.
Appeals to semantic compositionality accordingly take their place in the first instance in atomistic, bottom-up recursive explanations of the semantic interpretants of sentences in terms of the semantic intepretants of their sub-sentential components. It is true that both Davidson and Dummett (architects of two of the dominant approaches to the theory of meaning) privilege sentences in the order of semantic explanation—roughly because they are the minimal units that can be used to make a move in the language game, as Wittgenstein might have said. But they are immediately concerned to show how an initial grasp on the truth conditions of a finite set of basic sentences can be leveraged by a language-learner into mastery of the semantic contribution made by their constituent words to the meanings of an indefinitely large number of novel sentences. The explanatory target of such traditional compositional theories of meaning is the meaning and understanding of sentences. Hermeneutics, by contrast, focuses its attention to begin with on the meaning and understanding of suprasentential items such as texts and traditions. Hermeneutics is motivated by an appreciation of how surprisingly hard it can be to understand or grasp the meaning of a whole text, given that in some ordinary sense we may know perfectly well what each of its component sentences means in isolation. The dimension of meaning with which hermeneutics concerns itself is holistic and top-down. For it looks at the effect on our reading of one sentence that is produced by the other sentences that surround it—in many senses of ‘surround’. The semantic effect of context on meaning makes difficult what in some sense ought to be easy: putting together the meanings of separate sentences to get the meaning of a larger discourse. Phenomena quite different in kind from recursive compositionality seem to come into play in understanding, say, the contribution a particular sentence makes to the interpretation of a speech, novel, or poem in which it occurs. Making sense of novel texts is not just doing more of what we are doing anyway in making sense of novel sentences.
I have a general interest in seeing what these traditions have to teach one another. Given that my own work has been squarely in the tradition of theories of meaning, in this essay I will try to say something about what light I think at least one particular theory of meaning can shed on some classical claims and concerns of the hermeneutic tradition. My general answer to the question “What can hermeneutics teach us about theories of meaning?” is that it sets some important criteria of adequacy: hermeneutic theory describes semantic phenomena that any adequate theory of meaning ought to be able to explain. My general answer to the question “What can theories of meaning teach us about the problems of hermeneutics?” is that the theory of meaning ought to provide the conceptual raw materials for explaining the phenomena concerning the interpretation of texts that hermeneutics has pointed out. Theories of meaning ought, of course, to fill in the picture of sentential understanding that hermeneutics tends to take for granted in framing its distinctive concerns. But I think a proper theory of meaning can go much farther than this in making sense of hermeneutic phenomena. That is the explanatory enterprise to which the rest of my remarks here are intended to contribute.
One concern common to the two ways of approaching linguistic meaning and understanding is whether it is possible to separate what belongs to the meaning of linguistic expressions from the contribution to our understanding of them that is made by supplementing the meanings with imported external, that is, extra-semantic, in a broad sense, contextual considerations. A paradigmatic example of this concern on the side of theories of meaning is Quine’s argument that one cannot in principle distinguish language from theory, cannot isolate meanings from the effects of collateral beliefs. A paradigmatic example of this concern on the side of hermeneutic theories is the effect of interpreters’ traditions, historical situations, and cultural concerns on their readings of texts produced as contributions to quite different traditions, in different situations, addressing different concerns. In each case, the challenge is to say how, if meanings cannot be even theoretically insulated from effects by what is external to the meanings, we can be entitled to talk about meanings at all. And the threat is that if we cannot draw such a line, then “anything goes” in assigning meanings to sentences or assessing readings of texts. For what happens then to the notion of semantic constraint on what we make of a claim or a text?
If, for instance, as in Harman’s rendering of Quine’s semantic holism, we say that whenever I notice a cloud obscuring the sun, the meaning of all my words changes, how then can we think of understanding or grasping the meaning of those words at all? Again, if each reader is allowed to interpret a text in the light of her own concerns and convictions, what are the rules of the game? What sense can then be made of the difference between interpreting a text (finding a meaning in it) and just making something up (putting a meaning into it, or forcing a meaning on it)? One dimension constitutive of the space in which interpretation takes place is defined by the contrast between, at one extreme, what in jurisprudence are called “black letter” readings—which insist that each attributed claim be backed up by a sentence in the text that explicitly asserts it—and, at the other extreme, the hermeneutic ventriloquism practiced when the author’s lips move, but only the reader’s voice can be heard. This is catachresis: doing violence to the text, forcing one’s uninvited interpretive attentions on the unresisting textual corpses of the mighty dead. Both of the theoretical semantic traditions I am talking about acknowledge the importance of this challenge: to explain how the words that are uttered or appear on a page constrain how it is legitimate to understand or interpret them.
I think the best way to respond must begin by looking at the presuppositions of the challenge. The thought behind it is that the meanings of texts should be found and not made by interpreters. There is a way of thinking about meaning implicit in worrying about imposing a sense on a text, rather than discovering one, and it is part of a picture of which we should be suspicious. Enlightenment hermeneutics was thoroughly intentionalist. The author uses language as an instrument for the expression of thoughts that have the content they do independently of any such possibility of expression. Communication is successful if the ideas aroused in the reader have the same contents as those the speaker intended to elicit by those words. One’s task as audience is to take out of what is said the same crystalline, self-contained meaning that the author put in. Thus baldly put, hardly anyone would today subscribe to this Lockean picture—though some contemporary Griceans come close. And it is no better to recoil, with some contemporary neo-romantics, to the opposite extreme by seeing texts as shrinking to mere occasions for the imposition of meaning by their readers—as putting no constraints whatever on the free interpretive play of those who succeed in making them mean something (in the only sense in which anything ever means anything) by taking them to mean something. Such a view simply assigns to the audience the very same mythical meaning-constituting role the first view assigns to the author. Each is an unrecognizable version of the reciprocal relations of authority and responsibility that articulate the actual production and consumption of conceptual contents. The home language game of the making/finding distinction is empirical-practical discourse. There one clearly sees the two normative directions of fit Anscombe identified in her parable of the two grocery lists: the shopper’s authoritative for what groceries are correctly bought, the detective’s responsible for its correctness to what groceries are actually bought.2 It is by no means obvious that the making/finding distinction applies in anything like the same way to hermeneutic discourse, where the task is discursive understanding of episodes of concept application, acknowledgings of inferentially articulated commitments. 3
2. Hermeneutic platitudes
Gadamer4 has developed a hermeneutic idiom that articulates a via media between seeing a text as simply dictating the meaning to be found there, on the one hand, and seeing it as a tabula rasa on which readers are free to inscribe whatever meaning they wish, on the other. For him, meaning is not fixed by the contents of the intentional states of either authors or readers. Such states amount, in effect, simply to more text that is up for interpretation in the same sense as the text they are associated with. They can be considered, but doing so is just addressing a somewhat more capacious text than that with which one started. They provide just one sort of context within which a text can be understood. But there are others.
Another of his guiding ideas is that there is no such thing as the meaning of a text in isolation from its context—at least the context of its reading. A text can only be read from some point of view, in some context. The interpreter’s own attitudes and commitments form one such context. Meaning emerges in a process, which has the form of a dialogue in which the text is just one of the players. Meaning is a product of the words on the page and other features of the context in which it is situated—for instance, a tradition in which it features, or the concerns and questions a reader brings to the text. Understanding (practical grasp of meaning) consists in exercising a practical capacity to adjudicate the reciprocal claims of authority and responsibility on the part of the text and various contexts.
Relativizing assignments of meaning to contexts entails a pluralism about the meaning of texts. Texts can be assessed with respect to many different contexts and kind of context. Each provides a perspective on ‘the’ meaning. Or perhaps it is better just to talk about the sort of understanding that consists in being able to navigate with and among these perspectives.5 Further, the set of possible readings, contextual perspectives, is open-ended. There is no determinate totality of contexts. For each new text makes possible new contexts. This is one reason each generation, indeed, each reader, must reread and reinterpret potentially tradition-defining texts, and rethink the assimilations and affiliations by which they are put into the context of a tradition. The fact that I am addressing these issues as a philosopher, and directing my remarks to an audience of philosophers, brings specifically philosophical traditions to the fore. But the point is not limited to that kind of discursive inheritance. As T.S. Eliot said in his essay “Tradition and the Individual Talent”:
No poet, no artist of any art, has his complete meaning alone. His significance, his appreciation, is the appreciation of his relation to the dead poets and artists. You cannot value him alone; you must set him, for contrast and comparison, among the dead…
The necessity that he shall conform, that he shall cohere, is not one-sided; what happens when a new work of art is created is something that happens simultaneously to all the works of art which preceded it. The existing monuments form an ideal order among themselves, which is modified by the introduction of the new (the really new) work of art among them. The existing order is complete before the new work arrives; for order to persist after the supervention of novelty, the whole existing order must be, if ever so slightly, altered; and so the relations, proportions, values of each work of art toward the whole are readjusted; and this is conformity between the old and the new.
The denial of certain sorts of authority to the author of a text (avoiding what Foucault called “fetishizing the segmentation of discourse by signatures”), the relativization of meaning to context in a very broad sense, the model of dialogue, meaning pluralism, the open-endedness and mutability of semantic perspectives—I propose to call these by now familiar ways of talking “gadamerian platitudes”. By calling them that, I do not mean to impugn their originality or significance, but rather to mark that they have, thanks to Gadamer’s work, become platitudes expressing a select set of the framework attunements of hermeneutic theory.
Calling them ‘platitudes’ suggests I think we should believe them. But it is even more important to understand them. What is a context, and how does meaning emerge from putting a text into one? Talk of dialogue needs to be underwritten by an account of how each of the parties (text, context) exerts some sort of friction or non-determining constraint on the reading that emerges from their interaction—so that not just anything goes. The gadamerian platitudes are just the sort of thing it seems to me we should want to be entitled to say about the interpretation of texts. But earning the entitlement to the commitments those platitudes express requires real work. In particular, I want to claim, it requires a theory of meaning that can provide a model validating such hermeneutic truisms. Making sense of hermeneutic practice, as codified in the gadamerian platitudes, should be seen as a basic criterion of adequacy of a theory of meaning. And conversely, being interpretable in terms of an independently motivatable theory of meaning should serve as a basic criterion of adequacy of our hermeneutic practice. The principal philosopher who explicitly aimed for this sort of reflective equilibrium between his practice of interpreting philosophical texts and his theory of conceptual content is Hegel—which is one of the reasons I find him particular interesting. I want to indicate here how an inferentialist understanding of conceptual content of the sort I lay out in Making It Explicit underwrites and explains some of the axial gadamerian hermeneutic platitudes.
3. De dicto specifications of conceptual content
The target I’ll address here is just one kind of interpretation: grasping the conceptual content expressed by an utterance or text. Gadamer addresses a more general notion of interpretation, without the restriction to specifically conceptual understanding. On the inferentialist semantic conception, to be conceptually contentful in the most basic sense is to play a role as premise and conclusion in inferences.6 Conceptual content is understood as role in reasoning. The sort of understanding that is the aim of conceptual interpretation, then, is mastery of an inferential role: the ability to distinguish what follows from a claim, and what would be evidence for or against it, what one would be committing oneself to by asserting it, and what could entitle one to such a commitment. So on this semantic approach, it is the inferential roles of sentences that must be systematically generated by considering the contribution to the role of sentences as inferential premises and conclusions that is made by the occurrence of various subsentential expressions that occur in them. Although I cannot pursue the matter here, elsewhere I have shown how one can follow Frege in appealing to substitution inferences to accomplish this.7
The first most important sort of context for assessing the conceptual content of an utterance or text is, accordingly, its inferential context. For the inferential significance of a claim—what follows from it—depends on what other claims one can treat as auxiliary hypotheses in extracting those consequences. Different sets of collateral premises will yield different consequences. (This is the basis for the “Duhem point” Quine relies on in “Two Dogmas of Empiricism” to argue in effect from the claim that meaning must at least determine inferential role, to the holist claim that the unit of meaning must be no smaller than a whole theory.) If I already know the fruit is a raspberry, then being told that it is red will entitle me to conclude that it is ripe. But if instead I knew to begin with that the fruit is a blackberry, then being told that it is red will entitle me to conclude that it is not ripe. (As the child’s slogan has it: “Blackberries are red when they’re green, and black when they’re red.”) The inferential significance of the claim that the fruit is red depends on the context of background commitments with respect to which it is assessed. The material inferences that articulate the conceptual contents expressed by ordinary, nonlogical sentences are in general multipremise inferences.8 Each set of further premises with which a claim can be conjoined is a further context in which its inferential significance can be extracted and assessed.
Such a picture is not only consistent but comfortable with taking it that what really follows from any given set of premises can be a perfectly objective matter of fact. If the sample is copper and it is heated to 1083.4° C, then it will melt. In the same way, each text (in the minimally structured sense of a set of declarative sentences) has a definite inferential significance in each context of further claims. Abstractly, nothing privileges any of these contexts over any others; each highlights a genuine aspect of the overall inferential role played by that text, the contribution it makes to the goodness of inferences. Pragmatically, however, some contexts are privileged either by their relation to the circumstances of production of the text, or by their relation to the circumstances of its interpretation (and perhaps in other ways, too depending on the practices governing the inferential scorekeeping).9
One inferential context that provides a perspective on conceptual content of obvious importance consists in other things the author of a particular remark or text believed. Looking at the other commitments an author would acknowledge tells an interpreter something about what the author took it that she was committing herself to by making a certain claim, what she would have regarded as evidence for it or against it, and so on. So it tells us something about how she understood what she was claiming. Drawing the auxiliary hypotheses for extracting the inferential consequences of a claim from other commitments by the same author, or from the same work is one natural way to privilege a class of inferential contexts. When such an interpretation of a conceptual content is made explicit in an ascription of propositional attitude, it takes the form of a de dicto specification of the content of the attributed commitment. This is the basic form of indirect discourse, in which a claim that could have been quoted in direct discourse is instead paraphrased. In the same Eliot essay from which the passage quoted above is drawn we find:
Some one said: ‘The dead writers are remote from us because we know so much more than they did.’ Precisely, and they are that which we know.
But I can characterize his claim in terms he would presumably have found acceptable:
Eliot claims that we do know more than dead writers did, and that they are what we know.
The idea of de dicto specifications of conceptual content is for the ascriber to use words that in her mouth express the same content as the words that the target did or would have used express. There are some clear rules for paraphrases—especially those that involve other languages or indexicals. If Hegel says in German “Die Vernunft ist die Gewißheit des Bewußtseins, alle Realität zu sein,” I can ascribe that commitment in English: “Hegel says that reason is consciousness’ certainty of being all reality.” If you say “I am bewildered by Hegel’s claim,”10 I can ascribe the same commitment in my words by “You said that you are bewildered by Hegel’s claim.” Other standards of paraphrase are less clear cut. If someone claims both that Kant is a great philosopher and that Kant revered Hamann, we might attribute also the belief that a great philosopher revered Hamann, even though that particular claim had not explicitly been made. For it follows, by reasoning we expect the believer in question to accept, from the two commitments that were explicitly acknowledged. But what is one to do where the figure in question explicitly denies what appears to be a straightforward consequence of other commitments she avows? On the one hand, merely saying that one is not committed to something does not automatically mean that one is not, if it genuinely is a consequence of other commitments one has acknowledged. (It is not sufficient for an author to respond to the criticism that his views commit him to the objectionable consequence that p by pointing out that on page 97 he explicitly denies that p.) On the other hand, such a disavowal may signal that the author understands some of those claims differently (attributes to the sentences on the page different inferential roles) than the interpreter does. Under such circumstances the rules for de dicto specification of the conceptual content of another’s commitments are not clear.
Another dimension along which the notion of de dicto content specification is not well-defined concerns the exact boundaries of the inferential context one is allowed to appeal to in matching the inferential significance of the reporting sentence (which occurs inside the ‘that’ clause of the ascription) and of the reported one (the words the author did or would use in acknowledging the ascribed commitment). The idea of this sort of content specification is to extract the inferential consequences (dually, what would be evidence for the claim) of a claim made in the text by appealing only to collateral premises or auxiliary hypotheses that are co-acknowledged with that claim. If the boundaries of the text containing the claim being ascribed are themselves clear, and if the text can be considered as having no structure beyond being a set of claims, then an interpreter has a reasonably straightforward criterion to apply. (It still won’t be wholly straightforward, for there are a lot of things that won’t be explicitly said in such a text but that are fair game to appeal to in extracting the consequences of what is said: truisms such as that there have been black dogs, that freedom is better than slavery, that thorns can puncture the skin… . And issues can arise about the boundaries of the class of such truisms it is licit to invoke in particular cases.) But if the text in which the claim in question is made has further structure—for instance, a narrative structure—then complications arise. For instance: is it appropriate to appeal to claims made early in the narrative to interpret those made later? In Making It Explicit, for instance, the notion of an inferential role is introduced in Chapter Two as articulated into the circumstances under which it is appropriate to apply the expression, and the appropriate consequences of doing so. But in the next chapter this undifferentiated notion of propriety is further subdivided, in terms of commitments and entitlements. All the earlier statements then need to be re-interpreted retrospectively, as one distinguishes between circumstances that would commit one to apply an expression and those that would entitle one to do so, and consequences one becomes committed to by applying it and those one becomes entitled to by doing so. Again, in Chapter Six of that work the notion of substitution is appealed to in order to extend the inferentialist semantic approach from sentences to subsentential expressions. But we learn in Chapter Seven that the notions of inference and substitution presuppose (it is actually a reciprocal sense dependence relation) that of a token-recurrence structure. Everything said in the earlier chapter is implicitly to be re-read in terms of this later notion. So it need not be a straightforward matter to say what, within a single well-defined text, counts as co-acknowledged with a given claim.
And of course, the boundaries of the text one is reading can themselves be quite elastic. Ought we to worry about whether Hegel changed his mind about the structure and aim of the book he was writing halfway through the Phenomenology?11 Are we allowed to appeal to things he says in the Science of Logic in reading things he says in the Phenomenology? What about statements of Fichte’s with which he seems to agree? In reading Sein und Zeit, is it alright to appeal to what Heidegger says in his Grundprobleme, since that was written before his famous Kehre, but not to the Letter on Humanism, which was written afterwards? Different choices of context for de dicto ascription of conceptual content may have different virtues, provide different sorts of illumination. The beginning of responsible interpretation must be to make clear just how the boundaries of the context one is appealing to are determined—and so what the rules are for the sort of de dicto interpretation one is engaged in.
The motivating idea of de dicto specifications of the conceptual content of ascribed commitments is that the inferential context is to be supplied by the circumstances of production of the text. One engaged in this sort of interpretation is trying to specify the contents of commitments in a way that would be recognized and acknowledged as specifications of those contents by the one whose commitments they are. One is to take only the minimal account of the inevitable differences of doxastic perspective between the speaker/writer and the interpreter/ascriber required to deal with differences of language and of indexical situation. There is a way of writing the history of philosophy that aspires to this condition. One seeks to know so thoroughly what an author actually said, how his thought developed over his lifetime, what the rhetorical strategy of each work is and how it was understood by its author as fitting into the oeuvre, what his extra-philosophical concerns, attitudes, and experiences were that one can answer questions on his behalf in something like his own voice. One wants to be able to say what the author would in fact have said in response to various questions of clarification and extension. This is the point of view from which it is silly to try to interpret Hume if one only knows his distinctively philosophical antecedents and context—if one has not also read Gibbon and Adam Smith and so on. When I was first apprenticed in intellectual history, my mentor12 explained to me that one could not responsibly expect to understand what a thinker meant by a particular claim until and unless one had read everything that thinker had read. How else could one know what those words meant in his mouth—what contrasts he had in mind, what he took himself to be agreeing with, qualifying, or rejecting by saying that? De dicto intellectual history is a demanding discipline. Just having the requisite mastery over everything a philosopher actually wrote is a daunting undertaking for such prolific writers as Leibniz, Kant, Hegel, and Heidegger (though less so for those with more surveyable corpora, such as Spinoza and Frege), even before one has tried to master the traditions to which they owed allegiance and the milieus in which they lived and worked. I have heard specialized uses of the terms defined so that an expert is someone who knows a great deal about these things, but only a scholar is in a position responsibly to make negative existential claims about them all: "Wittgenstein nowhere says "Meaning is use" (though he does say things like "Don't look to the meaning, look to the use), "No-one before Hegel ever took explaining how one ought to do intellectual history as a criterion of adequacy on his theory of determinate conceptual content," and so on. But inferences that depend on premises of this sort are among those de dicto specifications of conceptual content aspire to capture.