blog.kennypearce.net

I'm not sure I understand the relevance of semantic internalism vs. externalism to the stated argument.

Why can't even an internalist object: "why can't I 'non-accidentally' use a word correctly if my subjective states are ordinarily caused by the correct conditions of application obtaining?"

The internalist thinks that 'the correct conditions of application' are determined by internal states, as a matter of meta-semantics. But they aren't thereby committed to the phenomenalist semantic view that the correct conditions of application are conditions of internal states. (Our internal states might instead determine that certain external states of the world will count as the truthmakers for our utterances.)

Hi Richard-

I think Berkeley takes his opponents' position to be something like this: words get their reference in virtue of the fact that the word excites a particular idea in the mind. Ideas are ideas of things in virtue of resembling the things they are ideas of. The idea is the meaning of the word.

There are three elements to this picture:

1. Semantic Atomism: To every meaningful use of a word, there corresponds some entity which is the 'meaning' of that word (in this particular usage).


2. Univocity: In any two meaningful uses of one and the same word, the meaning is the same.


3. Semantic Internalism: Meanings are internal to speakers.

Berkeley attacks (1) and (2) (primarily (1), since (2) is really just a simplifying assumption and people mostly don't actually believe that it applies universally), but accepts (3). Meanings for Berkeley are not entities, and they don't even have to be propositional or referential in nature - he is a sort of proto-speech-act theorist. But meanings still have to be internal. If the meanings are objective conditions, existing outside speakers, then they are simply not internal in the relevant sense.

Your suggestion in the last paragraph seems to be something like this: perhaps meaning is 'in the head' in the sense that the correct conditions of application are the conditions in which my internal states are such-and-such. You would deny that the specification of the correct conditions is exhausted by my internal states, but simply use the external states to pick out particular states of the world. As long as there are external states which obtain by physical/nomological necessity whenever these internal states obtain, then the external states can be part of the meaning of the term.

Is this what you mean? If it is, it sounds a lot like Putnam to me, so I would call it an externalist theory, but perhaps I have made a mistake of some sort?

Alternatively, you could try something more like Berkeley's opponents and say that the word refers to the internal state and the internal state refers to the external state. I'm not sure how this would work, though, because not all of the states we are talking about seem to be representational.

Hi Kenny, I feel like I'm missing something here.

I take it that "correctness conditions" are what contemporary philosophers would call intensions or "truth conditions". So the sentence "Grass is green" is applied correctly just in case grass is green. (And 'grass' applies correctly, in any given world w, to all and only things that are grass in w.)

So I understand #4 as claiming that intensions for plain language terms are "internal to speakers", in the radical sense that only internal states are included in the intension. We can't refer to anything outside of ourselves.

But that's clearly a far more radical claim than ordinary semantic internalism. Internalists merely claim that meaning is "in the head" in the weak sense than any two intrinsic duplicates have all the same thoughts and mean all the same things. So if my term 'grass' has an intension that includes various material biological objects, then any duplicate of me will have the very same intension associated with their term 'grass'. In either case, our internal states suffice to determine what "meaning" or intension is associated with our use of the term, but the intension itself may well include external objects. (If we understand intensions as functions from possible worlds to extensions, the extensions thus returned may well be sets of external, mind-independent, material objects.)

So understood, #4 simply doesn't follow from 1-3. It could be that we "consistently and non-accidentally" say true things about the external world. (We correctly say, "the grass is turning brown" only when the grass really is turning brown.) The obvious explanation for our reliability here is that we have sensory capacities that detect these features in the world. Our perceptions of grass being a certain colour are typically caused by the grass actually being that colour. No need to bring in twin earth.

What am I missing?

Right. The claim that "only internal states are included in the intension" is supposed to be a conclusion of the argument, not a premise. The argument is a little bit difficult and there isn't one single passage in which Berkeley clearly formulates it, and I'm not totally satisfied with my own formulation, but let's see if we can clarify a little bit. One of the questions that really interests me here is whether advances in our understanding of language tend to undermine any of Berkeley's arguments.

(4) is not supposed to follow directly from internalism. Note that on the 'Lockean' theory (in scare-quotes to reserve judgment as to whether it's an accurate representation of Locke's view) words signify ideas and ideas represent external objects. Words don't signify external objects directly. Rather, they signify "things only as they are known and perceived by us," i.e. our ideas of them.

Berkeley denies the existence of these sorts of ideas. (E.g. we have no idea of red in general, but only of particular red images, so the word 'red' cannot correspond to any one particular idea.) He attacks semantic atomism at length. (I get the name from Winkler's book; Berkeley doesn't have a name for the doctrine but formulates it explicitly in both the Introduction to PHK and in Alciphron 7.) So if there is no individual entity which is the meaning of each word, what is it's meaning? The picture is that there are supposed to be certain subjective conditions under which the word is correctly uttered, and this is all the 'meaning' there is - no 'idea', whether Platonic or Cartesian. The conditions have to be subjective because if they weren't our correct use of language would be accidental. All of our experience is necessarily 'subjective', and that's all we can use to determine whether to utter a word.

As I noted, this is a very strong sense of 'non-accidental' and most philosophers will probably want to object on this point: the correctness of our language may very well be 'accidental' in the relevant sense. As long as the connection between our subjective states and the external world has enough regularity, we'll still get it right enough of the time that we won't fall into radical skepticism or declare language to be impossible or anything.

Your primary objection, as I understand you, is to my association of this objection with semantic externalism. Let's take your definition of internalism: "any two intrinsic duplicates have all the same thoughts and mean all the same things." If I am a mind and it is logically possible for minds to exist disembodied (which Berkeley and his opponents held), then there could be an intrinsic duplicate of me which is disembodied and not in contact with any external world - perhaps he is being deceived by the evil demon. The subjective conditions in which he used words would be the same as the subjective conditions in which I use them. So he has the same level of linguistic competence as I do, and he means the same things by his words as I do, but, if physical object-talk refers to external material substances, he uses language incorrectly. Therefore, my correct usage of language is, in the relevant sense, accidental - I only get it right (if I get it right) because the world is nice to me.

An externalist can simply claim that my duplicate means different things by his words than I mean by mine.

Now, it might be possible for an internalist to claim that my duplicate lacks linguistic competence because he has a basic misunderstanding of the world. Since my basic beliefs about the world are right (even if only fallibly), I have a level of linguistic competence my duplicate lacks: for instance, I believe that when I have tree perceptions there are (generally) trees in the external world, and I'm right about this. My duplicate believes it, but is wrong. So he believes (wrongly) that it is appropriate to utter 'tree' in the presence of tree perceptions.

On Berkeley's view, the fact that it is correct to utter 'tree' in the presence of tree perceptions is supposed to constitute the meaning of 'tree' so that I can't be deceived about it. There is an epistemic argument running in parallel to the semantic one. I think you can run the two arguments independently, but there is a definite possibility that I am wrong about that.

Anyway, I don't want to outright deny that an internalist can make such a reply, but I think the reply is much more natural coming from an externalist. The internalist will have to explain in virtue of what my duplicate's words have the intension they do (or take it as primitive), whereas the externalist has a ready account.

I hope this makes things less rather than more confusing. I don't have this issue totally clear in my own mind as yet.

Yeah, that's helpful. I worry that premise #3 is conflating two different senses of "using language correctly". The case of my deluded duplicate brings this out. He shares my linguistic competence, and so 'uses language correctly' in the internal sense that he understands what his words mean, and how the world would have to be for his statements to be true. But his terms are applied incorrectly in the objective sense that many of his statements turn out to be untrue (because the world is not how he take it to be). He speaks of "grass" when there isn't really any such stuff around, etc. (Though note that this is more an epistemic problem than a semantic one. Again, he knows what 'grass' means; he's merely mistaken about how the world is.)

Now, in claiming that "common sense is correct", I take you to be claiming that we consistently say things that are true. This premise is merely contingently true. It is true of me, but not of my deluded duplicate. [N.B. If #3 is meant to claim that we're necessarily (or certainly) undeluded, then realists will simply reject this premise. So the argument only gets off the ground if we understand it as making the weaker claim that as it happens, we consistently say true things.]

But then #4 doesn't follow, because we can reliably speak truly even if we're speaking about things external to us. And this is a point that's unaffected by the internalism/externalism issue. That debate is about whether the external world plays a role in determining what it is that our words mean. But the objection doesn't rely on this. It merely points out that external facts F could reliably cause us to have appropriate internal responses R, which (either by itself or in combination with the external facts) constitutes a perception or belief that F is the case.

Note that the internalism debate concerns whether R intrinsically represents F, or whether it depends on which external facts obtain. But either way, so long as R-states are reliably caused by F-facts, then that's enough (according to the objection) for our thoughts to be consistently and reliably ("non-accidentally", in the ordinary sense) true.

(So, while you seemed to want to find some difference in linguistic competence between ourselves and our deluded duplicates, this is completely unnecessary according to the presented objection. Mere reliability is much easier to come by than strict immunity against error. As you originally said, I can "non-accidentally use a word correctly [so long as] my subjective states are ordinarily caused by the correct conditions of application obtaining." Even if this same linguistic behaviour would be "incorrect" in wildly different circumstances, that doesn't prevent it from being reliable in actual circumstances.)

I think this suggests that my current attempt to separate the semantic argument from the epistemic one may have been unsuccessful. I seem to have run together the claim that we know what or words mean and the claim that we know the way the world is. Thanks for pointing this out.