Monday, March 28, 2011

Two Views of the Analytic/Synthetic Distinction?

I just reread this post of mine from June, 2008: The Analytic-Synthetic Distinction and The Knowledge Argument. My approach to the distinction is quite different from the one I presented ten days ago, but perhaps these two views are complementary.

In the first post, I approach the distinction in epistemological terms. Analyticity and syntheticity are two different ways of testing the validity of the same propositions. They are different propositional attitudes--different knowledge relationships towards propositions. In my more recent post, I regard the distinction as a way of indicating different sorts of speech acts: one in which rules are stated and another in which predictions about the world are made.

On the face of it, these are distinct ways of looking at the distinction, though they don't seem so far apart from each other. I'll have to think more about it to see if one entails the other. For now, I have to say that both views make sense to me, and I think there is a connection.

Sunday, March 27, 2011

Szabo on Semantics and Pragmatics, Part I

For my own edification, I present the following review of a paper I'm reading called "The Distinction Between Semantics and Pragmatics" by Zoltan Gendler Szabo, which appears in The Oxford Handbook of Philosophy of Language (2006). (All page references are to this text unless otherwise noted).

I. Introduction

Szabo defines pragmatics as "the study of contexts of utterance, or more precisely, a study of the way context can influence our understanding of linguistic utterances" (363). Semantics, in turn, is presented as "the study of linguistic meaning, or more precisely, the study of the relation between linguistic expressions and their meanings" (363).

This might look simple enough, but I find it impenetrable. Presumably, what a linguistic expression means in some circumstance depends on context. Thus, following Szabo's characterization, pragmatics and semantics both study the relation between linguistic expressions and their meanings. Similarly, if the relation between an expression and its meaning relies on context, then semantics must involve a study of contexts, and so semantics must also be a study of the way context can influence our understanding of linguistic utterances. So semantics includes pragmatics and pragmatics includes semantics. It does not seem like Szabo has identified two distinct areas of study.

Szabo presents the semantic/pragmatic distinction as a way of understanding different types of disagreement: semantic, substantive, and contextual. If two people disagree about the definitions of words, then they have a semantic disagreement. For example, two people might argue about whether or not the Evening Star is a star. This disagreement could simply be a disagreement about what the word "star" means. That would be a semantic disagreement, and not a substantive one. Szabo isn't so clear when it comes to contextual disagreements, however. He refers to a situation in which one person says X="The table looks good here" and another says Y="The table looks terrible here." In this case, the people could be referring to different places or different tables, they could be using different standards of evaluation, they could be talking about different points of view, and they could even be performing different sorts of speech acts: one could be asserting a belief while the other could be making a joke. In this case, Szabo says, the disagreement is not a disagreement about what the words mean. He says that the disagreement can persist even if both people agree on what the words they are using mean.

This isn't clear. I'm not sure what sort of disagreement Szabo is trying to characterize. If I say X in earnest and you respond with Y in jest, we do not have a disagreement. I might think we do, but I would be mistaken. If we are talking about different tables or places, again, we do not have a disagreement, even though we might confusedly think we did. If we are talking about different points of view, we do not have a disagreement, either. The only disagreement could be if we are arguing over different standards for evaluating the appearance of the table. I suppose we might regard such a disagreement as contextual, but that seems like a strange way of putting it. Whatever we call it, it still seems substantive. I wouldn't say it's factual, but who says all substantive disagreements are factual? While there is no authoritative source for deciding on the correct standard of measurement here, and so no fact of the matter to be settled, we are still disagreeing about something of substance: namely, whether or not we should apply one standard over another. That has consequences outside of our discourse. That sort of disagreement determines where the table ends up getting placed.

Szabo hasn't clearly defined what a contextual disagreement is or how it could be distinguished from a semantic or a substantive one. Yet, there is a line to be drawn, Szabo says, and he says it is the line between semantics and pragmatics. Unfortunately, I don't think he's clearly argued for any sort of line at all.

At this point, Szabo has so far only introduced the problem, so we must read on to see how well he resolves this apparently confused situation. His next step is to discuss problems with some of the ways other people have tried to draw the line between semantics and pragmatics. He discusses three, which he calls the semiotic conception, the indexical conception, and the cognitivist conception.

II. The Semiotic Conception

Szabo presents and then discards a classic work in semiotics, Charles Morris' Foundations of the Theory of Signs (1938). Morris distinguishes between syntactics, semantics, and pragmatics. This tripartite distinction still exists today, though the sense of the distinction remains contentious. Morris characterizes it thus: Syntactics is the study of the relations between signs and signs; semantics is the study of the relations between signs and what they designate; pragmatics is the study of the relationship between signs and their interpreters. A sign is defined by a process of mediation, such that "something takes account of something else mediately, i.e. by means of a third something" (Morris 1938: 3-4; quoted on page 365). Szabo clarifies that semantics is not the study of all the relations between signs and disgnata, but only those relations that define the sign as such.

Szabo's criticism of Morris is wrong in many ways. First, he says Morris "assigns a rather narrow scope to semantics" (366), because according to Morris, semantics does not deal with the meanings of terms that do not designate anything (such as prepositions, quantifiers, logical connectives, and so on). This is not a persuasive objection. For one thing, there are ways of treating all of these signs as having designata. Logical connectives might designate mental operations, for example. Or we might say that they have no designata, and that they have meaning in so far as they are related to other words; thus, their meaning is a proper subject of syntactical analysis, and not semantics. Either way, Morris' view is not clearly problematic.

Szabo next claims that, on Morris' account, semantics does not account for the meaning of indexicals (like "this" and "that"), either, since we need to know context in order to determine what they designate. Here Szabo is inserting his own view of pragmatics, assuming that pragmatics alone is able to identify contextual relevance. But Morris does not distinguish pragmatics in that way. Rather, as Szabo apparently forgot, pragmatics is principally a study of the relationship between signs and their interpreters. Of course context is important for pragmatics, but it can also be relevant to semantics, too. So the fact that we need contextual information to identify the designatum of a sign does not mean it is no longer a semantic issue.

Szabo then says Morris includes too much in the scope of pragmatics: since much of human life is occupied with signs, Szabo says, pragmatics would amount to "a comprehensive theory of human interactions" (367). That's rather hyperbolic. Surely humans interact in many ways without signs. But, yes, much of our interaction does involve signs. How is that a problem for Morris' approach?

Next Szabo accuses Morris of confusing the role of signs in biology. According to Szabo, "Morris suggests that the concept of sign may prove as fundamental for the biological sciences as the concept of atom is for the physical ones" (367). This reminds me of a point Dawkins makes in The Selfish Gene: A biological description of human beings should include cultural artifacts, like clothes, glasses, and so on. Indeed, wouldn't it include language and signs in general? Why not suppose that a biological study of human beings would include an account of how people use/interpret signs? Morris is right. Yet, Szabo says the only sorts of signs that are of any interest to biologists are DNA, since they are information carriers. It looks like Szabo's completely missed the point. And, by the way, DNA aren't signs by Morris' definition. They don't operate as mediators in a semiotic process.

Szabo's final criticism is that Morris fails to recognize an asymmetry between semantics and pragmatics. Szabo says that, while semantics can operate "in relative ignorance" of pragmatics, pragmatics must take semantics into account. Morris (wrongly, according to Szabo) says that pragmatics abstracts from semantics just as semantics abstracts from pragmatics. But it's not so clear Morris is wrong. Szabo rather seems to be presenting a very misleading picture of what Morris actually says. Szabo says that, according to Morris, the study of pragmatics cannot even recognize that signs bear relations to particular designata, that we can only pursue pragmatics by ignoring the fact that signs bear relations to things. This cannot be what Morris means. There is a difference between abstracting from semantic relations and simply ignoring them. I don't think Morris is suggesting that pragmatics can be developed without any knowledge of semantics. I think Szabo might also be mistaken in supposing that semantics (as defined by Morris) can operate in relative ignorance of pragmatics. If semantics is the study of relations between signs and designata, and these relations are defined via an irreducibly triadic relationship between signs, persons, and designata, then there is no way you could have a robust semantics that remained ignorant of pragmatics, just as you cannot have a pragmatics that remained ignorant of semantics.

In sum, Szabo's rejection of Morris' approach has no discernible basis, other than confusion.

III. The Indexical Conception

Szabo now turns to Richard Montague, who proposes that pragmatics should be firmly rooted in the notions of truth and satisfaction. On this view, the difference between semantics and pragmatics is that the former treats truth-function without worrying about context, while the latter treats truth-function "relative to an interpretation and also to a context of use" (368). What this means, says Szabo, is that the need for pragmatics is nothing more or less than the need to account for indexicals. It seems, however, that this idea is more clearly present in Bar-Hillel, and not so explicit in the quotations Szabo provides from Montague. I'm not familiar with Montague, however, so I cannot make a forceful objection. Whether it is really Bar-Hillel or Montague who advocates it, the view on the table is that pragmatics comes into play only when we need contextual information to assign designata to indexical signs.

Szabo identifies some problems with this approach. The first is that it is unwieldy. It is too hard (perhaps impossible) to systematically treat all of the ways truth-functionality depends on context. The second problem is that Montague does not account for implicatures--that is, situations in which the literal meaning of an utterance is not what the speaker means to say. For example, an utterance of "The ham sandwich is getting restless" may, in ordinary situations, be used to refer to a person, and not a sandwich--for example, if it is used to inform a cook that a customer is losing patience. Yet, we run into trouble if we start treating ordinary descriptions as if they were indexicals. So the indexical conception of context-dependence seems incomplete.

On the face of it, I agree with Szabo's criticism of the indexical conception of pragmatics. I haven't read the rest of the paper yet, but I hope to read it and have the rest of my review up soon.

Friday, March 18, 2011

The Analytic/Synthetic Distinction

I've been discussing moral noncognitivism lately, and the topic turned to the question of whether or not mathematical truths are analytic. Analytic truth is truth by definition (or truth by virtue of meaning), whereas synthetic truth is truth in relation to what is the case. There's a lot of controversial history in philosophy over how to construe this distinciton, and whether or not we should even take the distinction seriously. Some say it is not a principled distinction at all. Much of the disagreement stems from different attitudes towards the very notions of meaning and truth.

I prefer a pragmatic view of the distinction, since I regard meaning and truth as matters of how sentences are used, and not as properties of sentences themselves. Thus, sentences are neither analytic nor synthetic, but they can be used to make analytic or synthetic assertions. One consequence of this view is that all sentences can be used to make both analytic and synthetic assertions.

The distinction is drawn thus: An assertion is analytic if it is properly taken as a definition or rule. An assertion is synthetic if it is taken as a statement about a possible state of affairs.

For example, take this common example of analytic truth: "All bachelors are unmarried." Since the term "bachelor" means "unmarried man," the definitions of the terms indicate that this sentence is analytic. We naturally suppose that the sentence itself is true by definition. Yet, the sentence doesn't say anything. Only people say things, and a person could use this sentence in both analytic and synthetic ways. If somebody is using it to state the rule that equates bachelorhood with being unmarried, then they are making an analytic assertion. If they are using the sentence to say something about what may or may not be the case, however, then they are making a synthetic assertion. For example, somebody could use it to assert that every living bachelor is unmarried. That is a statement about what is the case. If we then investigated whether or not every living bachelor was unmarried, we would be able to confirm or deny the truth of the claim.

Of course, since we regard it as a rule, we expect that we must find that every living bachelor is unmarried. How could we possibly find one that was unmarried? Yes, if we regard the sentence as a rule, then of course it is a rule. My point is that we don't have to regard it that way. The sentence itself does not force this reading. It's only a convention which defines the sentence as a rule. If we have a hard time understanding this, it's only because we are so accustomed to this convention.

I think anybody who takes the analytic/synthetic distinction seriously will agree that analytic truths are rules. What is true by definition is true as a rule. It should be agreed, then, that analytic statements are assertions of rules. They can be corroborated by appeal to some rule-defining authority, but not by appeal to what is the case--unless we have disagreement about what is an appropriate rule-defining authority. My view is unique only in claiming that what defines a rule as such is not a sentence, but a convention, and that sentences can be used in both conventional and unconventional ways. I don't think that's an unreasonable position to take in the philosophy of language.

It might be harder for some to accept in the philosophy of mathematics, but I think the same goes for mathematical truths, like "2 + 2 = 4." We can take such equations as analytic truths, in so far as we regard them as rules. But we don't have to regard them that way. We can also use "2+2=4" to express a synthetic truth: namely, that if you add two to two you will get four. This statement is about what will happen, and not about what is or is not a rule.

There is nothing about sentences or equations themselves which makes them rules. We only think of them as rules because we use them to state rules.

Thursday, March 17, 2011

Richard Carrier on Moral Realism

As I've noted before, Richard Carrier has a problem with definitions. His recent argument for moral realism has several awkward moments, like when he says being a moral realist means "being able to ontologically ground the existence of moral facts"--as if the mere presence of moral realists proved that moral realism was true! But it's his ludicrous discussion of "ought" which prompted me to blog.

He approaches the definition as a rational choice theorist might: "ought" means whatever we would do if we reasoned logically and knew all of the relevant facts. But he makes no mention of goals, as if any logical and knowledgeable person would do the same as any other in any given circumstance. Furthermore, we generally make "ought" statements without supposing knowledge of all of the relevant facts. "Oughts" might normally imply a "given what we know" qualification, but not a "given what we would know if we knew everything we would need to know to make a fully informed decision."

Carrier then makes the absurd statement that moral oughts are just like ordinary oughts, differing only in that they supersede all the others. How is that supposed to work? If what one ought to do is whatever a logical and all-knowing person would do, then what sort of imperative could supersede that? Moral oughts are apparently what we would do if we had more than all of the relevant facts!

We might try to be charitable, and suppose that Carrier was just stumbling on his way towards his main point, which is that moral oughts define "supreme values." But the problem remains. How are these to be distinguished from other values? I guess whatever are your most superlative values are your moral values. But how does this play out? If I'm wondering whether or not I should buy rolls when I go to the store, how do I decide if this is a moral imperative? Do I wonder if buying rolls is what I should do above all else? Do I evaluate the imperative in relation to all of my other obligations? Do I see that buying rolls is not the most important choice in my life, and thus see that it is not a moral ought? And what if the choice is whether or not I should give to a charity? If I don't think of giving to a charity as being something I should do above all else, then should I conclude that my giving money to a charity has no moral dimension?

Carrier's definition of "moral ought" doesn't work for me. Moral oughts are unique, but not in the way Carrier suggests. They are unique in the sense of the obligation, and not merely in its relative strength.

In any case, one person's moral values are not necessarily shared by anyone else. Surely people can disagree about moral values. But if moral realism is correct, then such disagreements are disagreements about matters of fact. Yet, the disagreeing parties don't necessarily disagree about the facts of the matter. They might just have different values (or differently weighted values). Carrier has no way around this. He tries to ground moral facts in social welfare, but his attempt is hyperbolic and insufficient. He says, "the Golden Rule, like fire and language and tools, was universally invented by all cultures because adhering to it is the only way to maintain social and psychological homeostasis" (emphasis added). Again in the spirit of charity, I'll suppose that Carrier really meant to say something like this: the golden rule can be interpreted as a variation on the tit-for-tat strategy for cooperation, and there is evidence that tit-for-tat is a relatively stable social strategy which appears in many forms (or, perhaps, under various permutations) across most, if not all, known civilizations. But what's the point? That our moral values should be means of maintaining a stable and harmonious society? That we should value the greater good over our personal interests? If that's the case, then Carrier is defining a moral imperative into his definition of moral imperatives. That's not an ontological ground for moral facts. It's a moral one.

Perhaps Carrier wants to say that we do define moral oughts in terms of the greater good, and not in terms of our personal interests. So then moral facts are defined as facts about what is best for society as a whole. We could debate that. But even if we allowed it, it doesn't give Carrier what he wants. To ontologically ground such facts, you would need some metric for measuring the prosperity of society at large, and you would need some way of showing that your metric was the right one. Carrier hasn't made one step in that problematic direction.

Carrier isn't close to being able to ontologically ground the existence of moral facts. I guess he's not a moral realist after all.

Tuesday, March 1, 2011

Sketches of a Non-Cognitivist Account of Morality

I recently discussed my reasons for thinking we should look for a non-cognitivist account of morality. Now I want to take some positive steps outlining what I think such an account might look like. I'm still working through all this, so I expect to stumble a bit here and there.

First, there might be a misconception that moral non-cognitivists don't believe in morality, or don't believe that moral statements have the sort of reason-giving power that people normally think. The concern is that non-cognitivism doesn't account for morality as it is commonly understood, and that it tries to undermine the common notion of morality. On the contrary, non-cognitivism (in my view) should fully account for everything philosophically and psychologically interesting about morality. Morality is taken as a given, as a process which we all observe and participate in, and which stands in need of psychological exploration and philosophical elucidation.

Another common concern is that non-cognitivism reduces moral statements to statements of personal taste. We can easily see the difference between statements like "The Beatles are great!" and "Abortion is immoral." When we make moral judgments or commandments, we are making statements with a special sort of force. We are putting pressure on other people to do (or not do) something.

We should not underestimate the importance of statements of personal taste, nor should we oversimplify the non-cognitivist's position on this matter. Statements like "The Beatles are great!" do have some reason-giving power. Expressions of taste can direct other people's actions. They are significant and they can motivate people, cause conflicts, and even lead to a strong sense of camaraderie.

Yet, statements of taste don't quite seem morally binding. At least, if they are, it is in a very weak sense. It's common, I think, to feel some small sense of moral indignation when our favorite musical groups, sports clubs, artists, or what have you are disrespected or unfavorably treated by friends, neighbors, in the press, and so on. We think that they deserve better, and we might be sad or frustrated that the world doesn't share our views. There are plenty of degrees of difference here--how we feel about music is probably much stronger than how we feel about flavors of ice cream--but there is presumably a spectrum, and characteristically moral statements might be at one end of it. I'm not sure that the difference between "The Beatles are great!" and "Abortion is immoral!" is so fundamental. The difference may be one of degree, though the difference is significant enough to warrant an explanation.

Our moral sense is intrinsically connected to our sense of personhood. When we make moral judgments, we are making judgments about what it means to be a person. For example, when we are morally outraged, we might say somebody is not a person, but a monster. Or we might ask, "What kind of person would do that?" Moral corruption is corruption of personhood. Exhibiting moral excellence is being a good person.

We also seem to define ourselves--who we are, the kind of person we are--by our tastes. So, again, questions of personhood can involve questions of taste as well as questions of moral integrity. These questions might not be so dissimilar.

Personhood is not a biological concept, but a normative one, though I expect there are biological constraints on what we could rationally consider a living person. I suspect that we have physiological mechanisms which motivate our interest in converging on a shared sense of personhood, too. We have a biological interest in harmonious normativity. We thus produce morality, which is a process of constructing and enforcing norms. We thereby seek out and foster dignity, which is the feeling of being a good person.

We don't respond well to seemingly arbitrary moral dictums, nor are we always inclined to listen to people who express moral disagreement with us. When we confront each other with contrary moral views, we either attempt to negotiate a shared understanding of what it means to be a person or we reach an impasse. In such cases, we might appeal to some particular moral system or principle which we believe serves as a measure of dignified action. Not just anything can count as a standard in that regard, but it is not the case that some measures are the right ones and some are the wrong ones. There is no standard by which we could conceivably measure our standards of personhood. (And if we did devise such a meta-standard, we would still lack a standard by which we could judge that one.)

We can make factual statements about norms, of course. I can say, "It is wrong to X," and (thanks to contextual implication) just mean that X is not permitted by some agreed upon ethical system. In this case, my moral judgment (if I am making one at all) is not specifically about X, but about following a particular ethical system. So, it is a fact that X is wrong according to some system, but it is not a fact that it is morally right or wrong to follow that system.

Against non-cognitivism, Russell Blackford says that, in some cases at least, people who say "X is morally wrong" mean that X is objectively forbidden, and he says that such statements are simply false. I think "X is objectively forbidden" is incoherent, unless it means that X is forbidden by some objectively known set of principles. In that case, "objective" just means, "capable of being recognized by any properly situated observer." Any properly situated observer can recognize the principles laid out in, say, the American Declaration of Independence, for example. So that could provide us with some kind of objective moral authority. Not an absolute one, of course, but the notion of an absolute moral authority seems incoherent.

People might appeal to some objectively recognizable moral authority to justify their moral judgments, but I do not think that the meaning of any moral judgment just is an appeal to some moral authority. If that were the case, then the judgment would not be that X was wrong per se, but that some particular authority should be followed. If my moral judgment is that X is wrong because it is wrong to go against Y, then my moral judgment is about Y, not X. And this is still a normative claim, not a factual one.

The premise here is that moral commandments and the like are of a logically different kind. They are not factual statements. They are more like promises, such as the promise we make when we say "I do" during a marriage ceremony. It has a great deal of significance and force--it is even contractually binding--but it is not a statement of fact. And what is a promise, but an act of setting some inter-personal condition on ourselves? Moral statements are like that, except instead of imposing inter-personal conditions on ourselves alone, we are imposing them on all persons as such.

There's a lot more to be said on this topic, but hopefully I've sketched out some interesting ideas. In sum, we can (and do) rely on facts in our assessments about what is or can be considered a person. Moral judgments can be based on factual knowledge. I do not claim that all talk of persons is normative or outside the scope of factual discourse. But the category of persons, as such, is normative. It does not reduce to biological or physiological categories. Furthermore, when we make moral judgments, commandments, or statements in general, we are constructing a notion of personhood which extends to all conceivable persons (though it may entail epistemic or other constraints; for example, I might think X-ing is immoral, but I might not say you were immoral for X-ing if you lacked the required knowledge base to recognize the salient features that made your behavior immoral.). We are not appealing to biological or physiological facts, even if what we are doing is informed by such facts. We are forging a sense of personhood, imposing it on others as well as on ourselves. That is the sense of moral prescriptions.