4. Interpretation as construction. From these considerations I conclude that the Collingwood picture must give an incorrect account of the meaning of a philosophical text. But I also believe it is in part a perception of the falsity of the Collingwood picture that has led to some of the strange and paradoxical things certain literary theorists have said in recent years. Roland Barthes, for example, was apparently prepared to say such things as that the author of Spinoza’s text is not really the historical Spinoza at all; rather, the text is a product not of the author but of language, together with the creative reader.30 Others, such as Derrida and Foucault, have formulated similar paradoxes, designed to cast skeptical doubt on the whole idea of a text’s having any determinate meaning whatever.
I think those who say such things may sometimes be trying to get at something true, especially insofar as they are counteracting romantic views according quasi-divine status to the author of literary texts (analogous fantasies have often been entertained about ‘great philosophers’). But I strongly disapprove of the sayings nonetheless. For unlike the brilliantly ironical remark that opens the Discourse on Method, reminding us of the need for modesty and an egalitarian spirit in conducting all philosophical argument, the paradoxes of these literary theorists in fact only obscure the truth, first by mixing up an important insight with the error it ought to remove, and then further making that insight harder to accept by self-conceitedly calling attention not to it but only to their own outrageous absurdity.31
In what sense, if any, is the author of a text constructed by interpreters? It is true that we find out what a text means only through thoughtfully interpreting the text. To do this we may need all our philosophical resources, and the interpretation may be a creative theoretical construction employing concepts and theories not available to the historical author. But the point to insist on is that if we are successful, then what we get at through the construction is what the text itself means, and hence precisely what its author meant. It is Kant, and not we, who created the meaningful texts; our function is to recover, understand and articulate that meaning. It may be true that Kant himself could not have fully understood or articulated his meaning. But we garble this truth if we mix it unawares with the Collingwood picture, and say that because in interpreting Kant’s text we aren’t reproducing (or even trying to reproduce) Kant’s actual thought-processes, therefore the meaning we are finding (or constructing) is not his meaning. Instead, what we should say is that because each new generation of philosophers devises new concepts in terms of which to ask questions and construct answers to them on behalf of Kant’s texts, the process of understanding better what those texts mean is forever ongoing. It will not end until people cease to read Kant, or at any rate until they cease to understand him.
Especially to be avoided is the (deplorably common, but crudely fallacious) inference from: “In discerning the meaning of a text, we need to engage in creative acts of intellectual construction” to “The meaning of the text is constructed by us, not put there by the author.” In general, when people find out facts about the real world (in the natural sciences, say, or in the study of history), they do so by constructing theories (about the origin of the solar system or the causes of the First World War). The fact that the theories are their intellectual constructs obviously does not entail that the external reality the theories are about (the solar system or the Great War) is not real but is only a figment of their minds. On the contrary, the whole point of these theories was from the start to grasp what is true about the real world as it is there independently of us.
In this respect, interpretation is no different from any other sort of theoretical inquiry about the real world. There are objective facts, independent of what we think, about what the text of the Critique of Pure Reason means (that is, about what Kant meant in the Critique). Contrary to what the Collingwood picture might suggest, these facts partly transcend anything Kant himself could have actually thought. We get at them through our intellectual constructs. When the constructs are successful, they tell us what Kant means in the text of the Critique.
5. Conversing with the dead. Many of the things I am trying to say here were expressed over a dozen years ago by Edwin Curley in an admirable article entitled “Dialogues with the Dead.”32 One insightful thing Curley said was: “Knowing what a philosopher means by what he says requires, at the very least, having some well-founded beliefs about how he would respond to questions and objections he may never have explicitly considered.”33 Curley’s basic thought here is one that has been famously stressed by Hans-Georg Gadamer: namely, that the meaning of a text is revealed only by asking it questions and understanding it as answers to them (though I don’t pretend that I am using that basic thought quite in the way Gadamer is). To the basic thought, Curley adds that the questions may, and even must, go beyond the questions the author explicitly asked in the text. Understanding a language requires being able to form original sentences in the language, expressing thoughts no one has ever had before. Likewise, understanding a text requires knowing, or at least having well-founded beliefs about, the answers the text gives to questions the text does not ask and which may never have been asked before. These questions must be our questions, simply because it is we who are interpreting the text and trying to understand it. Because the answers given by the text must be responsive to our questions, the meaning of the text must also be expressed in our concepts.
But I am not happy with one aspect of Curley’s formulation of this point. He speaks of “how [the philosopher] would respond to [our] questions…” This seems to me still too close to the Collingwood picture, since it too identifies the meaning of the text with the author’s thoughts, merely substituting counterfactual conditional claims about the author’s conjectured thoughts for past indicative claims about the author’s actual thoughts. This is connected with another thing Curley says, with which I also do not agree: “If our philosopher were a contemporary, still alive, active and cooperative, we might of course simply ask him what he means [by what he says in a text].”34 Here Curley writes as if our asking unasked questions and forming well-grounded conjectures applies only to the interpretation of dead philosophers, or at any rate those philosophers from whom, for one reason or another, we cannot elicit direct answers to our questions about what they mean. He seems to be saying that if only Aristotle or Spinoza or Kant were alive and willing to answer our questions about what he means, then that answer would necessarily be correct – absolutely definitive of the meaning of the text.
Yet Kant is right in saying that we sometimes understand a philosopher better than he understood himself;35 and this might be just as true of a living philosopher as of a dead one. Suppose I am asked what I meant by a statement in a philosophical essay I wrote five years ago and I give an answer. It is entirely possible that another person should reject my answer and propose an alternative interpretation of my statement. It is also entirely possible that she might be right and I might be wrong. No doubt in most cases, the author of a text is as likely as anyone to interpret his own statements correctly. We might also take the author’s interpretive statement as a further text, which as interpreters we must integrate into the construction through which we retrieve the meaning of the original text. If so, then this adds weight to the presumption that the author’s interpretation is correct. But even then the presumption is always rebuttable.36
Sometimes students ask me what Kant would say, if he were alive today, about some philosophical question we raise about his text. When they do, I always point out to them that Kant became senile several years before his death in 1804. Hence if by some ghoulish miracle of medical science he had been kept alive until today, then we would be lucky if he could even drool in response to our questions.37
This answer is, I think, entirely correct, and the fact that it doesn’t satisfy the students is simply a sign that they are asking the wrong question. When pressed, what they think they mean to ask is: “What would Kant say if he were brought back to life in full possession of his mature intellectual powers?” This new question, however, is both unclear and problematic. Are they or aren’t they also supposing that the resurrected Kant is aware of all the philosophical developments that have occurred in the last two hundred years (including the two centuries of Kant-interpretation and Kant-revision) which now shape the interpretive questions we are asking about his texts? If we don’t suppose that he is, then we cannot take it for granted that he could even understand many of our questions correctly, in which case his answers surely could not be taken as definitive of his meaning. But if we are imagining a Kant who is philosophically up to date (a twentieth century philosopher rather than an eighteenth century philosopher), then our question obviously has even more counterfactual suppositions than were apparent. It is not clear what sort of animal a twentieth century Kant would be, or whether this animal is any more thinkable than the legendary chimera or goat-stag, which philosophers in many ages have used as paradigmatic of the absurd and the unthinkable.38
6. Meaning as a norm for responses to questions. I submit that what we really mean to ask is not what Kant would answer to our questions about his text, but rather what answer would be given by an ideally intelligent and informed person who is also ideally intimate with the meaning of the Kantian texts. The notion that only Kant himself could perfectly satisfy this last condition probably reflects the baleful influence of the Collingwood picture, which equates such an intimacy with an awareness of what was passing through Kant’s mind as he was writing the Critique of Pure Reason; we are supposed to imagine that the resurrected Kant would answer our questions on the basis of a perfect memory of his two hundred year old mental processes, as well as a perfect knowledge of the last two hundred years of philosophy. It is a mistake to identify knowledge of the meaning of Kant’s text with the former sort of acquaintance; what I am suggesting is that the latter sort of acquaintance might turn out to be even more crucial to gaining knowledge of what his text means.
Once we free ourselves from the Collingwood picture, we can see that the right question to ask is not a question about what Kant would say when asked about the meaning of his texts, but simply a question about what we should say when we ask about them. The idea we have been examining, when reduced to these terms, might seem to be tautological, hence to get us nowhere. But this is not quite true. For it does make clear that knowing what a text means involves having justified beliefs about how someone (anyone) should respond to interpretive questions, based on the meaning of what the author said in the text. Or: for a text to mean something is for it to have somehow established a norm for such responses. When we interpret a text, what we are basically trying to do is articulate that norm. We do this by asking the text questions and trying to answer them in a manner which is determined by an intellectual construction we have devised on the basis of our understanding of the text.
Della Rocca’s interpretation of Spinoza, for example, arises from a set of questions, one of which is whether the apparent inconsistency between Spinoza’s theses (1)-(3) can be removed. Della Rocca’s interpretation articulates a norm for successfully answering that question (and a number of other ones besides). By drawing on twentieth century notions such as intensionality, referential opacity, and the mind-relativity of content, Della Rocca specifies a norm for answering such questions. The resulting answers are coherent. The norm is grounded in Spinoza’s text by referring to things Spinoza actually says, based on what his Latin words mean and on the background beliefs, assumptions and concerns it is reasonable to ascribe to him in light of the historical context in which the Ethics was written.
But what makes it the case that the constructed norm gives the right answers, relative to the text, even to questions the author never explicitly asked (and perhaps could not have asked)? Here I will attempt only a partial and tentative answer to this question. And I will place at least as much emphasis on some possible criteria the constructed norm does not have to satisfy as on desiderata the norm should try to meet. But I will try to say enough to shed some light on the curious fact that although interpretive disputes in the history of philosophy are sometimes endlessly controversial and frustratingly murky, they are nevertheless about something real and there are right and wrong answers to them.
I take it to be obvious, to begin with, that the norm does not have to yield correct answers to the philosophical questions we ask the text.39 To require that it do so would be to claim that what any text means must always be true and never false. But even the greatest philosophers are fallible human beings like the rest of us. I venture to say that in the writings of every philosopher, whether living or dead, who has written a significant amount about philosophical problems that are hard enough to be significant, there are already some assertions that we can know to be falsehoods. Hence I infer that the corresponding answers to questions the philosopher did not ask would therefore probably include some more evident falsehoods.
The first desideratum is that the norm should tell us to say all the things the text actually says, and gives its answers to all the questions it explicitly raises. We may be tempted to say that this is not merely a desideratum but an indispensable condition which any interpretation must meet. But that temptation is one which should be resisted. For the best interpretation of a text might be one that tells us that the author should not have said something the text does explicitly say. This is what Delahunty holds about Spinoza’s assertion that the mind and the body are the same thing. It is also what some hold about Marx and the justice of capitalist exploitation. All informed interpreters know that Marx explicitly held that capitalist distribution is not unjust. Some, however, think that these assertions must be set aside as inconsistent with Marx’s basic position. Their view, as G. A. Cohen once put it (with admirable candor), is that “Marx thought capitalism was unjust, but he did not think that he thought so.”40 Now as a matter of fact, I think that Delahunty is wrong about Spinoza and Cohen is wrong about Marx. But their interpretations might have turned out to be acceptable; indeed, they would even have been inescapable if philosophical tenacity and imagination had not been successful in devising coherent interpretations of apparently conflicting texts.
A second desideratum is that the norm should, all other things being equal, try to maximize the coherence of the responses it yields. This is a version of what is sometimes called the ‘principle of charity’. We should suppose, as far as possible, that what the text means makes consistent sense, and that this meaning yields answers to the various questions we ask the text that are at least consistent and mutually supportive, even if they are not always true. It is essential to realize, however, that all other things are often not equal. Sometimes they are not equal because (as Delahunty thinks about Spinoza, and Cohen about Marx) the text says things the author should not have said.
Things also may not be equal due to a third important desideratum: An interpretation of a philosopher should try to preserve and do justice to the philosopher’s most important and enduring philosophical insights. Sometimes a philosopher is very good at perceiving the pre-theoretical intuitions on a topic which need to be accommodated in a philosophical account, but less successful at constructing an account that accommodates all of them. John Locke seems to me a good example of a philosopher who combines this virtue with this failing. There are also great philosophers, such as I take Nietzsche to be, for whom systematic coherence, or even doctrinal consistency, is simply less important than other philosophical aims, such as creatively expanding the philosophical perspectives that we have available to us. If that is right, then any interpretation of Locke or Nietzsche that maximizes overall coherence will require us to ignore or exclude some of their most admirable insights, or at least to twist them or blunt their force. Sometimes the best interpretation of a philosopher is one which highlights the tensions or gaps in the philosopher’s views, and shows us how (and why) the philosopher is downright inconsistent.
A fourth desideratum is that the overall coherence should be not merely internal to the norm itself, but also coherence with background beliefs we have reason to think the philosopher held. It is chiefly in order to achieve an accurate idea of these expectations that we need linguistic mastery of the text, and erudition about its historical context and the background beliefs of its original intended audience. I have said that we must ask a text questions the philosopher could not have explicitly raised and our norm must sometimes give answers the philosopher lacked the conceptual vocabulary to give. But that is utterly different from saying that we may simply substitute our own beliefs for the author’s in determining the coherence every interpretive norm should seek. At the same time, the coherence must obviously be one that we are capable of grasping since we must regard the text as coherent. To that extent, our own philosophical beliefs inevitably play a role in determining it as well.
The task of interpreting a text is further complicated by the fact that sometimes we think a philosopher’s views have changed significantly over time. Should we attempt an interpretation of Aristotle on substance that reconciles the position of the Categories with that of Metaphysics , or should we decide that the two texts are irreconcilable and interpret each as advocating a distinctive position? Are Kant’s conceptions of reason and judgment the same in the third Critique as they were in the first Critique? Depending on which choice we make, what we say about what is meant in the Categories or the first Critique might be quite different.
Different interpretive enterprises may also have different aims in this regard: A detailed commentary on the Categories might interpret its statements about substance differently from a study that attempts a comprehensive interpretation of Aristotle’s philosophy as a whole. A commentary on Kant’s Groundwork must attempt to understand and expound the relation of morality to freedom in terms of the argument presented in the Third Section of that work, even though Kant seems to have abandoned that argument only three years later in the Critique of Practical Reason, and in the Religion he certainly rejects the Groundwork’s apparent equation of empirically motivated volition with the determination of the empirical self by a natural causal determinism. No attempt to give an encompassing account of Kant’s thoughts about freedom, morality and the determination of the will can afford to limit itself to an exposition of any one of these texts. Such an attempt has to undertake a sympathetic reconstruction of what the expositor takes to be Kant’s best thinking on this topic, as presented in a series of writings over many years. That may well result in an interpretation which does not perfectly agree with what any individual text says, and yet it might for all that be exactly the right interpretation of Kant’s theory of freedom.
7. Conflicting interpretations and the truth about what a dead philosopher means. I have given only a few desiderata we should try to satisfy in constructing an interpretation, regarded as a norm for answering questions. There may be still others, equally important, that I have overlooked.41 As I have said, I do not pretend that my account is complete. Perhaps it is not even as clear as it properly should be. But I think it gives us enough to make a couple of important points.
First, an interpretation that satisfies some of the desiderata – say, maximizing coherence, or taking best account of the background of meaning and belief we may draw from the historical context, may not do as well at satisfying others – say, giving due weight to what strikes us as the philosopher’s best insights. For this reason, it is often the case that no single interpretation is ideal. Often the benefits of an interpretation can be got only at significant costs (for example, it highlights original insights that are of direct interest to philosophers today, but forces you to repudiate some things the text says or to acknowledge that the philosopher’s view is not well supported by arguments in the text).
Second, how well a given interpretation satisfies the various desiderata (or even a single one of them) cannot be reduced to a set of rules or a decision procedure that guarantees a definite answer in every case. Which of two competing interpretations is to be preferred is nearly always a matter of judgment about the way a set of heterogenous desiderata bear on a specific and complicated configuration of particular facts. Interpreting a text well is the kind of thing Aristotle brought under the intellectual virtue of phronesis.
One obvious ground for disagreement about the meaning of a text will be differences between the interpreters’ own philosophical views. These may easily lead to opposing estimates of two rival interpretations regarding the value and importance of the alleged insights they capture, or their relative degree of coherence, or the relative plausibility of the philosophical theories they impute to the text. This makes it easier to see why disagreements over what a text means cannot be sharply separated from disagreements about which philosophical views are true and which are false, or which arguments are good and which bad, even though they are always disagreements about what a specific text means and not merely about what is philosophically true and false.
This point also helps us to understand why the interpretation of philosophical texts is likely to be a matter of endless controversy. For as long as philosophical questions themselves are matters of dispute (which will probably be forever), answers to them will also be subject to change regarding both their content and their evidentiary support. Accordingly, the strength of the evidence for a given interpretation of a certain historical text will vary in subtle ways with the perceived plausibility of varying philosophical viewpoints. Along with such variations, different interpretations of the same philosopher will become more or less defensible than they used to be. Further, the invention of new philosophical concepts and theories will make possible new interpretations, and also new versions of old interpretations that may be stronger than earlier versions were. The quest for what Aristotle or Spinoza or Kant meant will be endless, and these controversies will be, as they should be, deeply entangled with the ongoing collective quest to redefine and understand the philosophical issues with which Aristotle, Spinoza and Kant were grappling.