Sober discusses three main categories of metaethical views, which he calls subjectivism, conventionalism, and realism. These four chapters in a sense constitute a defense of realism, although the defense consists almost entirely of responses to arguments for the other views.
I. Subjectivism. This is the view that ethical statements are neither true nor false (or, as we might also put it, the view that there are no ethical facts). This view seems to be presupposed by the distinction between "fact and opinion" that is often taught in elementary schools.
We might distinguish between two subvarieties of subjectivism. First, one might hold that, since there are no ethical facts for ethical language to describe, we should just stop using ethical language altogether. This could be regarded as a version of nihilism. Or second, one might hold that ethical language has some useful function other than expressing facts. Some philosophers have held that ethical language is a means of expressing emotions rather than of stating facts, so that "x is good," for instance, means something like "yay x!" and "x is bad" means something like "x, yuck!" or "boo x!" This view is often called emotivism.
Sober considers, and rejects, a number of arguments for subjectivism:
A. Argument from disagreement:
1. There are ethical disagreements that can't be resolved.
2. If there were ethical facts, it would be possible to resolve ethical disagreements.
3. There are no ethical facts.
(Sober doesn't formulate this as an explicit argument, but this seems to capture what he has in mind.) Reply: the second premise is false. The existence of facts does not necessarily make disagreements resolvable. For example, creationists and evolutionists disagree about the origin of species in a way that shows no signs of finally being resolved, but there is surely a fact of the matter about which is correct.
It's also not clear that the first premise is true. Possibly ethical issues are resolvable, just difficult. And in fact there seems to be wider agreement on many ethical issues now than in the past; for example, it would be difficult now to find someone who defended the practice of slavery.
B. The genetic fallacy. (The fact that Sober labels this argument a fallacy gives a pretty good indication of how successful he thinks it is!) We can explain how people arrive at their ethical beliefs as a result of training (and, possibly, genetic predispositions). Therefore there are no ethical facts. ?? -- it's not clear exactly how the conclusion is supposed to follow, and indeed this is Sober's criticism. From the fact that we can explain where a set of beliefs came from, it doesn't follow that we have discredited the beliefs. (In the next chapter, though, Sober considers a more sophisticated version of this argument due to Harman.)
C. The Is/Ought Gap (Hume). Hume distinguished between 'is'-statements and 'ought'-statements, and argued that you cannot deduce the latter from the former. This might suggest something like this argument:
1. You can't deduce an 'ought'-statement from 'is'-statements.
2. Every true statement can be deduced from 'is'-statements (and so can the negation of every false statement).
3. 'ought'-statements are neither true nor false.
There's a lot of room for possible further refinements here. Notice that premise 2 is a kind of analog in the metaphysical realm to foundationalism in the epistemological realm.
The weak point here appears to be premise 2. One response is to point out that there are other means than deduction of arriving at one belief from others (notably abduction). A different response is to deny that we need to derive ethical statements from 'is'-statements by any sort of reasoning in order to make them respectable: that ethical statements are different from and not derivable from 'is'-statements does not necessarily make them nonfactual.
(Hume's distinction is often expressed as a distinction between fact and value, but this is question-begging, since it simply assumes that there can't be facts about values!)
[note: I'm passing over the argument of G. E. Moore that is sometimes referred to as his criticism of "the naturalistic fallacy."]
D. Explaining Ethical Observations. Gilbert Harman has argued in favor of a kind of subjectivism by arguing that there is a fundamental difference between ethics and science. Although ethics involves reasoning, and also involves observation, Harman thinks that we need to invoke physical facts to explain scientific observations, but we don't need to invoke ethical facts to explain ethical observations. Thus we have an abductive argument in favor of scientific facts, but we do not have a similar abductive argument for ethical facts. So simplicity suggests that we should reject the idea that there are ethical facts.
Sober's Response. Sober's rather brief response to this is essentially that, although ethical facts are not needed to explain ethical observations or beliefs, they are needed to explain "the ethical properties of specific actions." So there is still an abductive argument for ethical facts, it is just not an argument that begins with observations.
Criticism of Sober's Response. It seems to me that there are at least two problems with Sober's response. (1) Sober seems to be saying that we need general ethical facts to explain particular ethical facts. But it's still not clear what reason we have to think there are particular ethical facts in the first place -- and of course if there aren't any particular ethical facts, then there is also no need for general ethical facts to explain them! (2) Even if the argument successfully showed that there are ethical facts, Sober's concession that these facts do not explain ethical "observations" seems to leave it a mystery how we could ever know what the ethical facts are.
An Alternative Response to Harman. A different approach to Harman's argument might begin from the observation that Harman thinks his objection also shows that there are no color facts. All we need to explain color perception are physical facts (e.g. which wave lengths of light are absorbed and reflected by the surface of an object, the effects of light on the retina, and so on). It seems to me that this does not show that there are no facts about color; rather, what it shows is that colors just are complex physical properties of objects. Since colors are "reducible" to physical properties, to say that the physical properties cause our observations is entirely compatible with saying that colors cause our observations. I think perhaps we should say the same thing about ethical properties: that ultimately they will be "reducible" to the kinds of properties that can cause perceptions or beliefs.
There is a lot more to be said about the analogy between color properties and "secondary qualities" like color. How we divide the spectrum up into colors, and which ones are seen as similar to which others, seems to depend as much on the human perceptual apparatus as on the physical properties themselves. Perhaps we should think of colors as the physical properties that cause a certain perceptual response in humans, under ideal conditions of observation. A parallel view about ethics is sometimes called the "ideal observer" view of ethics: the idea that, for instance, right actions are actions that would be approved of by an observer who knew all the facts, did not have a personal interest at stake (since this can bias one unfairly), etc.
II. Conventionalism. (Note: this is Sober's term for the three theories he discusses under this heading; it is not a widely used term.)
Sober describes "conventionalism" as the view that there are ethical facts, but they are made true by someone's "say-so." The views are distinguished by whose say-so is crucial.
A. The Divine Command Theory. This is the idea that ethical facts are determined by God's commands: what makes something right is that God commands it. As Plato pointed out, this seems to have the consequence that God himself cannot be said to act rightly. God could command absolutely anything, and it would then be the right thing to do. One may be tempted to reply, "But God would never command us to make each other miserable, because that would be wrong and God would never act wrongly." But this response completely gives up the divine command theory, because it implies that God himself is guided by something other than his own command.
Another way to put this (Plato's question): are actions right because God commands them, or does God command them because they are right? Anyone who takes the latter view rejects the divine command theory.
B. Relativism. (A better term might be "cultural relativism" or "social relativism.") This is the idea that what is right is whatever society says is right. It faces problems similar to the divine command theory, only perhaps worse: it seems to make it impossible to say that societies can have mistaken norms or views about what is right.
C. Existentialism. The idea that what is right is whatever the individual decides is right. (It may be misleading to use the label "existentialism" for this view, since even if it accurately describes something existentialists believe, it is only a small part of their view.)
This view seems to me to be nearly incoherent. It seems to me that if we drop the idea that ethical requirements apply to more than one person, there is nothing left of the idea of ethics at all. (Wittgenstein famously argued that there could not be a "private language." I don't know about that, but I do think that the idea of a "private ethics" is incoherent!)
If you think that, other things being equal, it is wrong to torture babies; and if you think that this is not wrong because anyone says so, and that its wrongness is not relative to the conventions of some community or other; then you are an ethical realist, at least with respect to this particular moral requirement.
I think that most people have at least some leaning toward realism about at least some moral requirements. You may have been convinced by one argument or another that there are no objective and nonrelative moral facts, but if you ask yourself whether you can really believe that there is no fact about whether it is wrong to torture babies (subjectivism), or whether, although it is wrong in our culture, it could be right in a culture that endorsed it (relativism), or whether, although it is wrong for you, it would be right for someone who decided to regard it as right (what Sober calls "existentialism"), you will have a hard time answering "yes."
The main stumbling block to realism is simply puzzlement about what could possibly ground the objectivity of moral facts: what could make them true in a way that was completely objective and nonrelative?
Some possible approaches:
1. Start with objective facts about goodness and badness. Perhaps it is an objective fact that if something causes a person pain, it is bad for that person, and if it causes the person pleasure, it is good for that person. Then perhaps an action is right if it maximizes pleasure and minimizes pain. This is the approach of utilitarianism.
2. A different way of starting with objective facts about goodness and badness: perhaps x is good for y iff x enables y to thrive or flourish. (So it makes good objective sense to say that getting plenty of water and sunlight is good for a plant, for instance.) This is Aristotle's approach.
3. Perhaps ethical requirements are more like mathematics than like science: perhaps they are based, not on empirical facts, but on sheer logical consistency. Maybe acting immorally is always at bottom a result of a certain kind of inconsistency (roughly, not holding yourself and others to the same standards). This is Kant's approach.
4. Still another possibility: perhaps ethical facts are relative in a sense, but not to what any particular person says or decides. Maybe an action is wrong if it would be disapproved of, not by any actual observer, but by an ideal observer: someone who knew all the relevant facts, was not biased, etc. This is sometimes called the "ideal observer theory" of ethics.