Lecture: Language as thought and different species of ought

WARNING: Jotted down hastily, not yet cleaned up or fit for consumption.

Sellars wants to give us a naturalistic account of intentionality.

Logical behaviorism / philosophical behaviorism Def: the view that one can analyze without remainder intentional vocabulary / intentional concepts into purely behavior characterizations / dispositions to publicly observable behavior (specified in a non-intentional vocabulary).

Introduced in Empiricism in the Philosophy of Mind (9 years earlier), distinct from what he here calls logical behaviorism. Logical behaviorism refers to a view he attributes to Ryle. JB Watson and BF Skinner promoted this in psychology. Sellars never endorsed this because he saw this as being an application of instrumentalism in the philosophy of science.

Observable things, at least we know they exist. Theoretical things, we’ve got to make risky inferences to get to them. But we can also make observational mistakes. Not just “I thought it was a fox but it was a dog", but categorical observational mistakes. We can give some concept an observational role (e.g. declare that we can observe X’s) yet no X’s exist, i.e. no thing has such no thing with such circumstances of application and consequences of application. E.g. we can have a theory of acids. "Anything that’s sour is an acid. And anything that’s an acid will turn litmus paper red." Well, then we have observational access to acids. If we eventually find something that tastes sour that turns litmus paper blue, then it turns out there are no acids, even though we could observe them (or: had every reason to believe we could). Likewise: being a witch was observable (even though there are no such things).

The Plasticity of Mind is about bad theories incorporating observational practices, i.e. "What do you mean there are no K’s. I can see K’s, there’s one right there!"

So again, this is a response to someone saying we can distinguish theoretical from observable entities by pointing to the fact that we can make mistakes about whole categories of theoretical entities.

Just because our evidence for attributing mental states comes from behavior does not mean, unless you are an instrumentalist, that you have to be able to define intentional concepts in terms of behavior. (This doesn’t mean that the intentional states are less real, just that we aren’t in a position to observe anything but the behavior)

• Digressions

  1. Semantics is a field with instrumentalist vs theoretical realist views. Michael Dummett is instrumentalist by observing the fact that meaning something is only understood through verbal behavior and concluding that any theory of meaning must be definable in terms of behavior. (A theoretical realist might postulate meanings as theoretical entities to explain verbal behavior and say our access to meanings is inferential and, if they are good theories, then verbal behavior gives us inferential access to something (meanings) that exist.)

  2. MacDowell and Sellars agree (and disagree with almost all others) that what you hear when someone talks to you is the words themselves, rather than hearing noises and (by some inferential process) constructing the words. You have to actually actively do some work to hear that mere noises. This is evidenced by how difficult it was to tell computers how to recognize a smile in a picture. (Some say it’s a contradiction to say that meanings are essentially normative yet, on the other hand, we sometimes can directly perceive them. But there’s nothing in principle unobservable about normative states of affairs - see Sellars’ criteria of observation below)

(Controversial) Criteria for observation:

1. You have the capacity to reliably and differentially respond to some normative state of affairs

2. You have to have the concept and which is a matter of inferential articulation and practical mastery of inferential proprieties, involving it. And then if you can hook the one up to the other, you’ve turned what was beforehand a theoretical concept for you into into the concept of an observable

Sellars wants to make sense of the notion of “language as a rule-governed enterprise" (as essentially involving norms). Sellars believes that if your account language doesn’t involven norms, you will be describing the vehicles by which we communicate, rather than what we’re saying/meaning.

Reminder that, due to the regress argument, that we need to broaden our notion of ‘rule’ from just explicit rules and need think of rules also as implicit in what we do. Sellars wants to better understand the relationship between implicit practical abilities and explicit representations of rules.

The question of whether meaning is a normative concept was brought to philosophical attention by Kripkenstein [1]. In present literature, Hattiangadi and Katherin Glüer have pushed back upon the idea that it is a normative concept, advanced by Brandom and MacDowell. Brandom feels it is because they haven’t learned lessons from Sellars, in particular thinking of norms purely in terms of explicit presecriptions and not making the distinction between ought-to-be’s and ought-to-do’s.

“You can define possibility in terms of not and necessity. You can define necessity in terms of not and possibility. I think it’s the beginning of wisdom to think of defining not in terms of the relationship between possibility and necessity, but I’m the only human being who thinks that."

Grice on non-natural meaning: reduces what a linguistic expression \(P\) means in terms of the meanings of thoughts and beliefs of those uttering \(P\). Sellars isn’t satisfied with this: the puzzling phenomena of meaning are common to both thought and language.

Sellars says “ought-to-be’s imply ought-to-do’s" but is not exact about what quantifier: all or some? Brandom thinks ‘some’ makes more sense, since there could be an ought-to-be requiring a state of affairs to change without telling us who has to do what to fix it (you need auxillary hypotheses to turn it into an ought-to-do). E.g. “all clocks should be in sync”.

With a trainer, someone with concepts/rules can condition language learners to shape their behavior (teach them ought-to-be’s). It’s important that it’s possible for the language enterprise get off the ground (i.e. without trainers). It’s possible for some sort of selection process to naturally reinforce ought-to-be’s (can be social but the conditioners need not be doing so intentionally).

We can deliberate making a distinction between ought-to-be’s in the context of humans vs nonliving/nonsentient beings (e.g. “plants ought to get enough water”). Ruth Millikan’s work relevant. Connects to the Aristotelian account.

Consider ought to be’s in the context of training animals: These rats ought to be in state \(\phi\) whenever \(\psi\).

• Could be just for rats, qua rats

  • they ought to be eating when they’re hungry, or something like that

• this could be something we want the rats to do

  • when they come to a branch in a maze, the rats go to the side that’s painted blue and not to the side that’s painted red.

  • That’s a regularity that ought to be not because we can read it off of the fundamental teleology of rats

  • The conformity of the rats in question to this rule does not require that they have a concept \(C\), e.g. of colors blue and red. We just require them to respond properly certain to differences emanating from \(C\). This doesn’t require even consciousness (photocells can respond differentially to colors).

“Recognitional capacity” gets systematically used in two fundamentally different senses (an ‘accordion word’)

• reliable, differential response

• applying a concept

Important for Sellars that following an ought-to-be requires only the former sense.

We should talk about learning a language as ‘coming into the language’ rather than ‘learning a language’. It’s more like the way one comes into a city. You come to be able to take part in an ongoing practice, as opposed to getting some intellectual insight.

Teaching the very young child to say ‘purple’ when showing her a purple lolipop is getting her to follow an ought-to-be just like the rat example. There is an ought-to-do for teachers of a language that they see to it that children produce the appropriate responses. This presupposes that the teachers do have a conceptual framework of ‘purple’ and of ‘vocalize‘ and what it is for an action to be called by a circumstance. The learner is not required to have any of these concepts. The ought-to-be is explicit in the teacher’s mind.

All of our evidence in science comes from empirical observation, so all of our concepts (and claims) must be translatable without remainder into observational vocabulary.

What warrant would there be for any conceptual excess beyond the language of our evidence?

There aren’t really any theoretical entities. We postulate them merely to characterize regularities of observable entities. Statements in the observational language are simply true or false, whereas statements in theoretical language are merely more or less useful.

This is a view that members of the Vienna Circle flirted with. A permanent temptation of the empiricist tradition.

The alternative is called theoretical realism. Sellars says this is a mistake (an example of "nothing-but-ism", along with emotivism in ethics), originating from thinking of the difference between observational vocabulary and theoretical vocabulary as an ontological difference in the objects referred to by those theories. But it’s not an ontological distinction, it’s a methodological distinction. Two different epistemic relations we can stand in to things that there are. Observable things are those that we come to know about by observation (non-inferential observation reports). Theoretical concepts are concepts that we can only be entitled to apply as a result of a process of inference. See Pluto example.

Linked by

• Language as thought and communication: Sellars

• A distinction fundamental for both ‘must’ in the alethic and doxastic modal senses.

• Sellars: You can’t understand either of these kinds of oughts without understanding both. In particular, if you try to do everything with ought-to-do’s:

  • one would fall into a kind of Cartesianism: we’d need to think of linguistic episodes as essentially the sort of thing brought about by an agent whose conceptualizing is not linguistic.

  • We’d be precluded from explaining what it means to have concepts in terms of the rules of the language. Ought to do’s have the form of “in circumstances \(C\), do \(A\)” - what language are \(C\) and \(A\) stated in? Regress of rules without ought-to-be’s.

• This is important because natural way to think of rules is exlusively in terms of Ought to Do (Sellars himself advocated this earlier: “A rule is always a rule for doing something"[2]).

• There is also an analogous distinction involving permission, rather than obligation.

You can’t understand what it is for somebody to be saying (and therefore thinking something) apart from the way they’re treated by some community. That’s a sense dependence: you can’t understand the one without the other.

It doesn’t follow that somebody who could do all of this wouldn’t have thoughts until/unless they were treated as having them; that would be reference dependence.

So just as an unconnected example, illustrating that distinction: suppose I defined ‘beautiful’ as “would cause pleasure in someone”. Now, then I’ve instituted a sense dependence between beauty and that sense and pleasure; if you can’t understand the concept of pleasure, you can’t understand the concept of beauty, which is a response dependent dependently defined consequence of it.

But now we ask, would there still be beauty if there were no pleasure? Were there beautiful sunset sunsets before there were any people to feel pleasure? That would be the reference dependence question. We say sure because they would have caused pleasure (if there were anyone there to feel it). And we can say, in a possible world in which there never were humans, it still could be that if there were, they would have responded to the sunsets with pleasure.

So, we could say there’s a sense dependence between these concepts, but there doesn’t need to be a reference dependence between doesn’t mean you can’t have the one without having the other it just means you can’t understand what one of them is without understanding the other.

So the claim would be that’s the relation between the thoughts and our normative attitudes are social attitudes. It’s not that the thoughts pop into existence at that point for them.

People with romantic views of language: Derrida and Nietzsche. These are compatible with Wittgenstein believing language has no downtown.

Sellars disagrees: the language-language inferential transitions are of the first importance among those because what makes the entries and exits language entries and exits is the way they connect to the inferential moves. And so he would say to Derrida, “yes, we do all of these other wonderful things with the language, but that’s all parasitic on the meanings that things are given because of the role they play in the space of reasons... now, once you’ve got that up and running, once you’ve got those meanings to work with, now you can start to do other playful things with it, e.g. use them metaphorically. All sorts of things become possible. But that’s in principle a superstructure on this structure."

Sellars story of how ‘the light dawns slowly over the whole’.

Both the infant and Koko the gorilla can be trained into a language (in the form of conforming to ought-to-be’s). At some point the human makes a jump - they have the concept and can be a trainer of others. What’s the nature of that jump?

For Sellars, this is a change in normative status, not a lightbulb that went off in one’s head. Like the change on your 21st birthday, when suddenly doing the very same thing, making the same pen scratches that you could have made the day before, would not be obliging yourself to pay the bank a certain amount of money every every month for the next 30 years. But after your 21st birthday, when you scratch your pen in exactly the same (physically descriptively, matter-of-factually) way, all of a sudden it has a hugely different normative significance because now you will be held responsible. You’ll be taken to have undertaken commitment in a way in which you were not eligible to undertake that commitment by doing the very same thing descriptively, the day before.

When you get good enough at the language game moves, you do get acknowledged by the community. We don’t characterize this physically-descriptively because we’re not describing someone / some matter-of-factual boundary that has been crossed. We’re not describing the child, we’re placing the child in the space of reasons.

It’s the difference between the one and a half year old, who toddles in to the living room. And as her first full sentence says, “Daddy, the house is on fire." Well, one doesn’t think that she has claimed that the house is on fire. She’s managed to put these words together, this is good. If the four year old comes into the living room and says “Daddy, the house was on fire", you hold her responsible, you say “how do you know? Did you smell smoke? And you know, what should we be doing? What follows if the house is on fire? What should we be doing?" You take her to have claimed this to have undertaken a commitment and you hold her responsible for it. The difference is not some light that’s going on. It’s a difference in normative status, ultimately a difference in social status.

This is the difference between just conforming to the pattern, and actually making claims. The radically anti-Cartesian aspect of Sellars is that this is also the the difference between conforming to the pattern and having thoughts at all.

However, as Dennett points out: you can treat any even inanimate object as an intentional system, e.g. this table as having the one desire that remain at the center of the universe. And the one belief that it is currently at the center of the universe, which is why it resists us moving it. (by extension, we treat our cats and dogs this way). So we should only treat things as thinking if we have to. Brandom takes an opposite view, that you should always treat something as talking if you can (note this is a very high bar).

The period prior to the child’s mastery and social status as a language speaker has some peculiarities. His verbal behavior would express his thoughts but, to put it paradoxically, the child could not express them. The child isn’t in a position to intentionally say that things are thus-and-so, even though it is in a position to say that things are thus-and-so. So there’s a question: which comes first, speaker’s meaning or semantic meaning?

Semantic meaning is a matter of what the words mean. No agent involved in that. In English, the word ‘molybdenum’ means the noble metal with 42 protons. Contrast with “When Humpty Dumpty says ‘glory’, he means a nice knockdown, drag-out fight”. Grice says speaker meaning comes first. Sellars says that is a Cartesian way of thinking about things, that the primary meaning is what words mean in the language process.

If I claim the notebook is made out of copper, I have (whether I know it or not), committed it to melting at 1084 oC and that it conducts electricity. My words mean those things, whether or not I mean to.

The kid produces vocal (not yet verbal) noises until he is a member of the language community (his verbal noises conform to enough ought-to-be’s).

As soon as he can say something, that’s the expression of a thought. To take him to be saying is to be taking him to be thinking out loud. It’s a further stage, when he can take expressing that thought as the object of an intention, and intentionally do as an action that say, before that, that’s just an act, it’s a performance, he can reliably produce appropriately, but not yet intentionally produce. An adult could be in this situation: Auction example. That’s the sort of position that the kid (who’s just crossed the line into being able to say something) is: she can produce a vocalization that will hold her responsible for, and which, accordingly, we take to express a thought. But she doesn’t yet have the concept. So, she can have the concept of its being red or the house being on fire. But not yet, the concept of endorsing something, or of making a claim that he’s saying can be a later development. And you need that concept in order to intend to be making a claim.

Important to make distinctions between different types of saying:

• mere utterance (position of 1 year old)

• saying that things are thus-and-so

  • having mastered the entries/exits/language-language moves, but no metalinguistic concepts

  • Could be called “merely thinking out loud”

  • Can perform speech acts.

  • Can express that something is read or even a desire for something (“I’m taking that’)

• intentionally saying (telling someone) that things are thus-and-so

  • need concepts of asserting/believing as well as concepts of thus-and-so

  • Can perform speech actions.

  • Self consciousness.

We need to think of the child as being able to give evidence without the concept of evidence. This is important in the story of how the language game gets off the ground with the early hominids. But we have real experience with this: when teaching logic, it’s helpful to teach students to have the practical mastery of writing proofs (prior to them having the concept of a proof). They first get familiar with the symbol pushing game. (proof is a strong form of evidence). This is very common in mathematics education.