The Scorn of Humpty Dumpty (Part 3)
In the previous entries in this series, I argued that:
(1) Huemer claims there are only three possible antirealist positions: noncognitivism, subjectivism, and error theory.
(2) I argued that this is false. And that this is false because each of these position appeals to the presumption that moral claims share a uniform and determinate meaning that commits speakers to just one of these three accounts.
(3) If we reject the UD assumption, we are given two additional possibilities: metaethical pluralism, and metaethical indeterminacy.
(4) It is difficult to assess the viability of pluralism and indeterminacy without a deeper examination of what we’re trying to do when describing moral claims. Both center on the meaning of ordinary moral claims - the claims made by nonphilosophers. One approach emphasizes speaker intent, and focuses on the mental states of the people making moral claims. We can call this an “internal” approach and it describes the metaethical stances of people making moral claims. The second approach is an “external” approach that attempts to offer an adequate account of the meaning of moral claims without necessarily purporting to capture any particular intentions or psychological states of the speakers, but instead captures what we could call metaethical commitments, which capture the meaning of moral claims without regard for the psychological states of speakers.
(5) This distinction presents traditional metaethics with a dilemma: if the goal is the former (capturing internal states / stances), such accounts are empirical, and the empirical evidence supports pluralism/indeterminacy. If it’s the latter, it relies on a contestable account of language and meaning that we are free to reject (and which I do reject) that, once rejected, opens the door to pluralism and indeterminacy. Of course, one could reject the dilemma and point to some third account, or one could embrace the dilemma, and argue that the external approach is the correct and appropriate one and that, if it’s correct, the three traditional antirealist positions are the only possible positions. Perhaps some contemporary analytic philosophers think those battles have already been fought and won. “Ordinary language philosophy is dead,” I’ve been told. I’m not convinced.
Let us now turn to why Huemer believes that noncognitivism, subjectivism, and error theory are the only possible metaethical positions. According to Huemer:
“Either ‘good’ purports to refer to a property—that is, ‘x is good’ asserts that x has a certain property—or it does not. If it does, then either some things have the property or nothing does, and either the property depends on observers or it doesn’t. If we say ‘good’ doesn’t purport to refer to a property at all, then we have ethical non-cognitivism. If we say ‘good’ purports to refer to a property but nothing has that property, then we have nihilism. If we say ‘good’ purports to refer to a property, and some things have that property, but the property depends on observers, then we have subjectivism. Lastly, if we say ‘good’ purports to purports to refer to a property, some things have that property, and the property does not depend on observers, then we have moral realism. Those are the only possibilities” (p. 5).
These may be the only possibilities if we accept the notion that the word “good” has a single uniform and determinate meaning, such that “good” either expresses a proposition or doesn’t, and either refers to stance-dependent properties or doesn’t. But I think the correct answer is “none of the above” because I don’t think words like “good” have any particular meaning, much less a uniform and determinate one.
And, in any case, to the extent that ordinary people use language like “good,” in everyday contexts, I don’t think the goal of such terms has anything to do with referring to stance-independent or stance-dependent facts. Rather, I believe such language only obtains meaning in virtue of the local, sociofunctional uses to which it is deployed, and that such language can serve a variety of goals, from persuading others, to signaling one’s desirable character traits, to making one’s rivals look bad, to indicating one’s alliances, to feeling better about oneself.
While one motivation among the many uses of normative moral language may be to report what the speaker takes to be moral facts, it’s unclear whether people have stable, rich metaphysical positions such that we can determinately point to such language as an indication that the speaker is referring to stance-independent or stance-dependent facts. Speakers may lack the relevant psychological states for us to say that they have a stance one way or the other, meaning that they have no determinate metaethical stance, and there may be no observable differences between an external description of their language as committing them to any particular form of realism or antirealism.
In other words, when someone says, “murder is wrong,” …if we were to examine their psychological states, the speaker themselves simply isn’t intending to describe stance-independent or stance-dependent facts, yielding metaethical indeterminacy with respect to this distinction.
And at the same time, if we were to examine their communicative goal in making this utterance, there may be no decisive evidence one way or the other that would allow us to conclude, with confidence, that their usage in this particular context best accords with any specific form of realism or antirealism.
With respect to the former, analytic philosophy simply isn’t equipped with the appropriate methods for resolving such disputes. Huemer and other philosophers don’t typically conduct empirical studies, and I’m not sure they think empirical studies would have much value.
This suggests that their methods are likely intended to yield a priori knowledge via the tools that are available to philosophers, such as conceptual analysis. I have many critical things to say about such armchair theorizing that I don’t even know where to start (at least not here), so I’ll focus on an observation on the methods typically employed. Philosophers like Huemer tend to rely on the analysis of toy sentences. That is, they don’t actually analyze actual moral utterances, but imaginary, or hypothetical utterances that a person could potentially employ in some real-world context. Their goal is to then analyze these toy sentences in order to assess what they take them to mean, by consulting their own intuitions and linguistic competence. For example, they might look at a sentence like
“Murder is wrong,”
And think about what the sentence means. It sure looks, at first glance, like a propositional claim. And it sure seems like we can embed such utterances in arguments, reason employing such utterances, and so on. For instance, consider the following argument:
P1: If assassination is an act of murder, then assasination is morally wrong.
P2: Assassination is an act of murder.
C: Therefore, assassination is morally wrong.
By consulting our intuitions or linguistic knowledge, nothing seems off about this argument. If moral utterances were merely expressions of emotions, as some noncognitivists suggest, then embedding moral utterances into an argumentative structure should prompt the intuition that something has gone very wrong. “Murder? Boo!” isn’t the sort of thing we could embed in an argument without yielding a very strange result that hardly makes sense. So, we might reason, moral utterances can’t merely express emotions, because they can figure into arguments.
Yet note that none of this analysis engages in any way with what ordinary people are actually trying to do when they say “murder is wrong” in the real world. All the armchair analysis in the world might give the impression to the philosopher that moral claims must be propositional because we can embed them in arguments, but suppose we went out and observed all actual instances in which an ordinary person used the term “murder is wrong,” and we found that they never actually did embed such utterances in arguments. In fact, they only ever used such utterances to express emotions, and even said so when asked.
“Yes, of course I only mean to express my opposition to murder!”
I’m not saying this is true. I’m quite confident it’s not. My only point is to ask you to consider if it were true. In other words, suppose 100% of all real-world instances in which a nonphilosopher uttered “murder is wrong,” were accompanied by the intention of the speaker merely to express an emotion, and never to assert any kind of propositional claim.
Would we still want to insist that, nevertheless, “murder is wrong” just is a propositional claim, and that what it means just is an attempt to describe a realm of stance-independent moral facts?
This would be extraordinarily weird, to say the least. What would it even mean to say “murder is wrong” is a propositional claim? Which instances of “murder is wrong” are we referring to? Would we insist that all the people going around intending to express their emotions by saying “murder is wrong” are really making propositional claims? I suppose we could do that, but if philosophers wanted to insist on this, I’d wonder what their motivation was. This would be a very strange way to think about language and meaning.
Of course, a philosopher may object that they agree, but that they aren’t intending to ignore the intentions of speakers. Rather, as competent speakers of a given language, with rigorous training in analyzing their own usage, they are making inferences about what nonphilosophers are intending to do when making such utterances. Even if all the nonphilosophers claimed to be expressing emotions but not expressing propositions, they lack the training, introspection, expertise, and perhaps the sheer talent of the philosopher to recognize that this just isn’t so. They are unfortunately self-deceived or incompetent or confused. We must turn to philosophers to tell us what we really mean.
If this sounds a bit dubious, that’s because it is. I doubt many philosophers would shamelessly make such a claim. Yet at times they do seem to come close. Responses to critiques from experimental philosophers (who survey nonphilosophers in the way I describe here) often insist that philosophers have special expertise in analyzing concepts, and that we can trust the philosopher and ignore laypeople’s judgments as unreliable. Merely aggregating a bunch of nonexperts is hardly going to yield an adequate account of what we really mean, after all, any more than we can understand quantum mechanics or evolutionary biology by surveying the public. We must turn to our esteemed analytic philosophers.
Perhaps philosophers really do obtain this kind of expertise. I very much doubt it. Perhaps this sounds to you, as it does to me, far more like the claims of mystics who insist they’ve been inducted into the Eleusinian mysteries and have sacred knowledge that can only be obtained by secret Mycenean rites which according to rumor involve the sacrifice of the goat and submerging oneself in an amphora full of garum.
If you want to know what people mean when they say things, there are fields better equipped to address such questions, such as linguistics and anthropology. It’s embarrassing that philosophers think they can figure out what everyone else means when they say such-and-such by sitting merely thinking about it on college campuses.
We already know surveys of thousands of people aren’t adequate to make generalizations to humanity as a whole, because the populations psychologists tend to draw on tend to be WEIRD populations (Western, Educated, Industrialized, Rich, and Democratic). Yet somehow we’re supposed to take an extraordinarily idiosyncratic, mostly-WEIRD population, most of whom are trained almost exclusively in academic traditions that are themselves the epitome of WEIRD, and suppose that these people’s appeals to how they think yields generalizable insights into how humanity as a whole thinks?
This makes about as much sense as a group of mountain climbers informally polling only those who arrive at the peaks of Mount Everest on whether “it’s satisfying to climb mountains,” obtaining an affirmative from everyone around them, and declaring that all people everywhere feel the same. It might even be more ridiculous than this, since people from all over the world climb mountains.
In any case, contemporary analytic philosophers rely primarily on the analysis of, as I’ve said, toy sentences. Why are they toy sentences? Because they are not actual sentences. Actual sentences are sentences actual people use in actual contexts. For some reason, philosophers seem to struggle to recognize the difference between the two. Take the utterance
“Murder is wrong.”
Note that this is not an actual utterance. If this isn’t obvious, ask yourself: who is uttering it?
The answer is: nobody. This isn’t an actual sentence. By an actual sentence, I mean an actual instance of a real person making use of such a sentence in a real-world context. For comparison, suppose you were a biologist that studied cats. Suppose someone said, “What do you study?”
Imagine the following responses one could give:
(a) Responding by saying, “Cats, like lions and tigers.”
(b) Showing the person a photograph of a cat.
(c) Pointing to an actual tiger, and saying “cats, like this one.”
When I talk about actual sentences, I mean (c). “A cat” isn’t a cat. It’s words someone would use to refer to cats. A picture of a cat is also not a cat. Only an actual cat is a cat. Just the same, “murder is wrong,” isn’t an actual moral utterance, it’s a kind of hypothetical example of the kind of utterance a person might actually use.
Toy sentences are not like actual sentences. Toy sentences exist outside any actual context of usage. They exist only in a kind of fictional Platonic playground that removes all of the elements of the contextually-embedded nature of a real-world instance of uttering “murder is wrong.” Isolated in this way, philosophers examine what such sentences “mean,” while ignoring (either unintentionally or by design) all the contextual factors that may be relevant to assessing such claims.
This is a mistake. Utterances are a type of behavior, and behavior only makes sense in the contexts in which it occurs. Take an instance of one person shoving another person, or hurling a rock in someone’s direction. Do these sound like violent, unethical actions? They may, if only because you’re imagining the most salient or typical instances in which a person may perform such actions that come to mind. Yet it is possible that, in a given context, someone is shoving someone to move them out of the way of a car, or hurling a rock in someone’s direction to get their attention and indicate that a danger is approaching. We couldn’t assess behaviors outside a context; we can only assess them by imputing a context onto an imaginary or toy scenario, such as
“One person shoving another person.”
In order to adequately analyze a “scenario” like this, we simply need more information. Why are they shoving the other person? What is their goal?
Now imagine you were asked to analyze one person shoving another person without any context. You might think, “all else being equal, that would be wrong.” Such judgments involve imagining a kind of quasi-context around such scenarios. One might think of one person shoving another person for fun, or for no particularly good reason at all, and then judge that this would be bad, since it would presumably cause harm to the other person, or upset them. But note that to make such judgments, we must fill in a lot of “blank spaces” in the underspecified context. We have to imagine, for instance, that the person being shoved could be harmed, or doesn’t want to be shoved, or… and so on.
Philosophers are constantly dealing with contextually impoverished pseudoutterances and pseudoscenarios that lack the contextual details of real world events. Absent that context, such events may be literally unanalyzable. They’re not even real events. With imputed context, they become analyzable, but the context imputed on them can vary from one judge to another, and, when it goes unstated, it is hard to know whether we’re imagining the same scenarios or different scenarios. Finally, much of the context that may be relevant to assessing issues in the real world is simply missing from such scenarios. Philosophers train, over many years, to think about such toy sentences and toy scenarios, and produce judgments about them. Presumably they imagine such judgments generalize to real world instances, but
More generally, I don’t think that facts about the existence of stance-independent moral facts turn on the contingent features of the way any particular human population is disposed to think or speak. I am not alone. Kahane (2013) maintains that moral realism does not require a semantic claim. As such, I don’t think that my rejection of moral realism requires a specific thesis about the meaning of ordinary moral claims.
In other words, all three of the traditional antirealist positions do seem to presuppose that moral claims have some shared meaning. I reject all three both because I think people making moral claims mean things when they make moral claims, but the claims themselves don’t mean anything apart from what people intend to mean when they make moral claims. And, second, because I think that people don’t generally intend to mean anything that comports with traditional metaethical theories.
I endorse folk metaethical indeterminacy: I don’t think ordinary people, when making moral claims, intend to express claims that determinately fit with traditional metaethical analyses, for much the same reason that I don’t think people’s causal claims determinately commit them to views about how to interpret quantum mechanics. I just don’t think ordinary moral claims are in the business of functioning exclusively or primarily to express propositional claims that make reference to stance-independent moral facts, or to report subjective preferences, or to express imperatives, and so on; they may on occasion serve in any or all of these roles, or none of them, depending on the context. And such contexts are so rich and varied that I find it absurd that any particular account could claim some type of semantic primacy over the rest, such that we could say that one particular account of what people are doing when they make moral claims is “correct,” and the rest are deviant or aberrant or parasitic, as Gill (2008) would put it.
Huemer talks about whether ‘good’ purports to refer to a property or not. But given my views of language, I’m not sure this makes sense, if interpreted literally. That is, interpreted literally, Huemer seems to be implying that the word good itself purports. Unfortunately, I’m not sure what Huemer means by “purport,” since “purport” can mean a variety of things: “intends,” “appears,” “claims,” and so on. “Appears” or “Superficially appears” doesn’t make sense of other uses of “purport” in the surrounding passages, since Huemer states that “Non-cognitivism holds that evaluative predicates do not even purportedly refer to any sort of property” (p. 4). “Appear” isn’t a plausible interpretation of this use of “purport” since noncognitivists probably don’t think that moral claims appear, or superficially appear, to be imperatives or nonpropositional emotional expressions. After all, the whole point of noncognitivism is typically that this is all they really are in spite of appearing to be propositional because they are often expressed in declarative sentences. Noncognitivists point to the motivational features of moral claims, or their social, expressive, or pragmatic roles as indications of lacking propositionality, but they don’t typically deny that the surface structure of moral claims looks (albeit perhaps superficially) like propositional claims. I don’t know that any noncognitivist (though perhaps there are exceptions) that deny that that such statements such as “murder is wrong” are prototypical moral utterances, and that such statements at least bear a structural resemblance to “fire is hot,” even if they aren’t used in the same way.
More importantly, noncognitivism isn’t a theory about what it looks like moral claims mean, it’s a theory about what they actually mean, so framing this as them appearing to not be propositional wouldn’t be an adequate characterization of noncognitivism: the view isn’t that they look nonpropositional, but that they are nonpropositional. Unless Huemer’s meaning of “purport” shifts across usages (which would be confusing for other reasons), I take this to hint at his meaning more closely reflecting the notion that moral claims themselves intend or claim things. Perhaps the view that sentences have meanings apart from the intentions of their speakers is popular among philosophers, but I take it to be one of the most obviously ludicrous notions in all of philosophy, and perhaps the central fulcrum around which the central errors of analytic philosophy center: the mistaken view that words and sentences mean things.
If this is what Huemer means, as I’ve suggested earlier, I don’t think that is how words or language work. On my view, words can’t purport to do anything, and sentences can’t claim or intend anything. Words and sentences don’t have intentions and don’t have any meaning outside a context of usage. Only people or agents can have intentions, mean things, and assert things. Sentences themselves can’t assert or mean things. For comparison, imagine we observed someone eating a sandwich. Would we say that the person is eating the sandwich? I sure would. But the way philosophers speak about language, it almost seems like they’d say “the act of chewing is eating the sandwich.” This is analogous to how philosophers often seem to view language: as though the words and sentences are themselves asserting and intending, almost as though they’re anthropomorphized. Human agents speaking to one another take a backseat when communicating. While we are using words, it’s not we who intend or mean or purport to say or express anything. Our words do. Humans speaking to one another become a mere epiphenomenon for the Platonic exchange of words with other words. The whole notion is fanciful and bizarre, so much so that when framed in what are admittedly and obviously critical terms, it sounds deeply uncharitable. Nevertheless, I stand by the suspicion that Huemer’s framing of the range of metaethical views available is so restrictive because Huemer is employing a conception of language that I don’t share (because I find it absurd).
Rather, I would side with Humpty Dumpty, who concisely conveys what I take to be a far better understanding of the way language works than it would seem many contemporary analytic philosophers do. In a passage from Through the Looking Glass, Humpty Dumpty has the following exchange with Alice
‘When I use a word,’ Humpty Dumpty said in a rather scornful tone, ‘it means just what I choose it to mean—neither more nor less.’ ‘The question is,’ said Alice,’ whether you can make words mean so many different things.’ (Carroll, 2005, p. 60)
Humpty Dumpty goes on to disagree, asking instead whether we control our words, or do our words control us? Humpty Dumpty presumably believes the former: we decide what we mean, and that’s all there is to the meaning of our words. Nothing more. Nothing less.
I agree with Humpty Dumpty. One could articulate this in the following maxim:
Words don’t mean things, people mean things.
References
Baz, A. (2012) When words are called for. Cambridge, MA. Harvard University Press.
Carroll, L. (2005). Through the looking Glass. In P. M. Parker (Eds.), Webster’s thesaurus edition for PSAT, SAT, GRE, LSAT, GMAT, and AP English Test Preparation. San Diego, CA: ICON Group International, Inc.
Kahane, G. (2013). Must metaethical realism make a semantic claim?. Journal of moral philosophy, 10(2), 148-178.
Loeb, D. (2008). Moral incoherentism: How to pull a metaphysical rabbit out of a semantic hat. In W. Sinnott-Armstrong (Ed.), Moral psychology: The cognitive science of morality (Vol. 2, pp. 355-386). Cambridge, MA: MIT Press.
Sinnott-Armstrong, W. (2009). Mixed-up meta-ethics. Philosophical Issues 19(1), 235-256.