Monday, June 19, 2006

An Illustration of the Link Between Verification and Uncertainty in Meaning

Suppose we assert the proposition "bananas are green, brown or yellow."

Suppose we then discover a natural banana tree that has mauve banana's, but which produces toxic fruit. We may have difficulty deciding whether our proposition still stands. Is such a toxic banana still a banana? Did we mean that only edible bananas are green, brown or yellow? Has our proposition been falsified?

I think that precision in meaning is generally quite good when our proposition has a particular verificationist intent. If we were programming a robot to collect bananas for humans to eat, toxic bananas would not be bananas (edible fruits) for our intended purposes. In that case, our original proposition stands undefeated by the mauve fruits.

On the other hand, if we were categorizing plant species, we might consider toxic mauve bananas to be bananas in good standing* (by morphology, genetic similarity, potential for cross-breeding, etc.), thereby falsifying our original claim.

In these two cases, the intended mechanisms of verification make the clarification of meaning fairly straightforward.

However, things become much more shaky when our goal is extension of meaning beyond verification. If we don't explicitly specify how our proposition should be verified, the meaning becomes more uncertain. Mauve, toxic bananas either are or are not defeaters of the proposition depending on our verification protocols. The blanket statement that bananas are green, brown or yellow is highly uncertain without knowing precisely how we would verify the claim.

*One doesn't often use "bananas in good standing" in a sentence!


Robin Zebrowski said...

This seems to me intrinsically flawed though. What about this mauve toxic fruit makes it a banana? Just the shape alone? Clearly shape doesn't qualify, or plantains are bananas, and I don't see any possible way that something with the usual chemical make up of a banana can be toxic and still *the same thing*. (Meaning, if it's got the same chemical makeup, it isn't toxic, and if it is toxic, it won't have the same chemical makeup).

This is just the problem posed by Hilary Putnam concerning natural kind terms in his Twin Earth experiments (with water-as-H2O and water-as-XYZ). I adore Putnam, but I've never gotten past the fact that we're made of H2O, and a chemically distinct product is simply not going to perform the same way in the real world.

It's always the real world getting in the way, isn't it?

Doctor Logic said...

Hi Robin,

I don't see any possible way that something with the usual chemical make up of a banana can be toxic and still *the same thing*.

You're saying that bananas are defined as having the same chemical makeup. Yet, it seems to me that this would lead you to exclude some banana species due to minor chemical differences (bananas were defined prior to knowledge of chemistry). As you survey known kinds of bananas, you would be continuously redefining your chemical criteria. When you discovered a new kind of banana, you have to decide whether it was a banana because of shape or taste (and you should widen your chemical criteria), or whether it should be excluded for not meeting the chemical constraints of previous types of bananas.

Also, it seems possible to me that a banana with a couple of extra genes could be purple and inedible, while perserving 99.9% of its chemical composition.

Upon finding such a plant, a botanist might conclude that the mauve musa was a banana in the same way that a giant panda is a bear.

Personally, I'm not sure what to make of the Twin Earth experiments. Semantic externalism doesn't seem to add any enlightenment. Yes, causal histories are vital, but no, water does not simultaneously mean H2O and XYZ. It stands for experiences (causal histories) of H2O and XYZ, which is not the same thing.

Actually, I thought of your work when I wrote this post. If meaning is inextricably linked to experiential verification, that's another way of seeing that embodiment is necessary for realistic AI's to understand our meaning.

Robin Zebrowski said...

Yeah, I knew I'd get in trouble by going the chemical composition route :) It was the first thing that came to mind, and as soon as I said it I realized I was basically committing all the errors of the Twin Earth experiments all over again. I think the real point I wanted to make was that for any given definition of banana, you'd be unlikely to find one toxic to us that still fits the definition of banana (and perhaps it's for experiential reasons, or perhaps its for stronger definitional reasons, but I think you're right that it's the former). I think keeping in mind the problem of continuation of species and the fact that the real world is really ugly when it comes to "natural kinds" just complicates things further.

Maybe I can make that my job - I can complicate things, but I can't solve them!

Ken Brown said...

Hey doctor(logic), I'm still enjoying your posts even though I don't comment very often. I've posted a response at Signs of the Times.

Colin Caret said...

Doc, you said bananas were defined prior to knowledge of chemistry

I don't see how this is relevant to the meaning externalism that Robin was trying to motivate. It is also true that water was labeled before he advent of Daltonian chemistry, but it is also true that water is H20, a fact about water that we discovered after the advent of Daltonian chemistry.

Couldn't the same be true of bananas? And if it is true that there is some micro-structural essence of bananas qua natural kind category, then isn't the verificationist stuff a moot point? It would now just be a question whether this mauve fruit falls into that natural kind category, which is just a question about whether it has the right essential features, the right micro-structural properties.

Of course, the problem with all of this is not that it fails for bananas alone, but that it seems to fail for biological kinds in general. For instance, it is highly implausible that the reason we group bananas as one thing and plantains as another thing is because of significant chemical differences. Likewise, we do not distinguish Cardinals and Bluejays based on their microstructural differences. If there are any such differences, they tend to be things which are irrelevant to our grouping stuff that way.

Biology was always the area where Putnamian meaning externalism was the least plausible. However, I still do not see just why verificationism is any more plausible. When I say "I like bananas", what kind of verificationis content does that sentence have exactly?

Doctor Logic said...

Hi Colin,

I'm not sure I'm following you.

I'm saying that when we say "I like bananas," we're saying so for verificationist reasons, e.g., because we like their shape, their taste, their color or whatever. Different people, might have different reasons for saying they like bananas. So, the stage may be set for a schism within the World Banana Party, should mauve bananas be discovered. Though all the party members subscribed to the claim "I like bananas," they each meant something slightly different by that statement when they joined (e.g., color, taste, consistency, consistency when cooked, shape, etc.). The imprecision in each member's verificationist meaning opened the door to schism when bananas could be found that would fail to meet their experiential reasons for liking bananas as they originally defined them.

Thus, priority is important. Water was defined by our pre-chemistry experiences of it (transparent, liquid, tasteless, thirst-quenching, etc.). Finding that water is H2O just correlates our prior experience of water with other phenomena (e.g., electrolysis). It would be wrong to conclude that the Romans meant H2O when they discussed the construction of aquaducts. They knew nothing of chemistry, so the liquid to which they referred was possibly independent of chemical structure. If we discover an XYZ that is different from H2O, but which has the same experiential qualities (given Roman technological abilities), then the Romans meant XYZ as much as they meant H2O. We would simply have correlated our experience of water with both H2O and XYZ. (Heavy-Oxygen Water might work as an XYZ.) Indeed, it's possible (though ridiculously unlikely), that water on Earth was different 1500 years ago than it is now. Then the Romans may actually have been referring to XYZ. Yet, we and the Romans mean the same thing when talk of water independently of its fine structure. We would only mean different things if we made statements about water in chemical reactions that depend on differences between H2O and XYZ.

If I have completely misunderstood your question, please feel free to tell me so. :)

Colin Caret said...

Well, I think that stuff about XYZ is wrong, but its sort of beside the point. Let me focus on what you said at the beginning of your comment...

I'm saying that when we say "I like bananas," we're saying so for verificationist reasons, e.g., because we like their shape, their taste, their color or whatever

Now, this to me is a very weird claim. I agree that we all might assert "I like bananas" for different reasons. Some of these might be reasons like enjoying the color, taste, etc. Another reason might be an emotional attachment to the banana's place in the food chain. Or perhaps I like bananas because I have a fetish for the chemical and nutritional structure of bananas. All of these are clearly reasons that one might like bananas, but I don't see what verificationism has to do with it. Can you explain the connection?

Doctor Logic said...


Let's take any reason you might think of for liking bananas, e.g., your suggestion:

Or perhaps I like bananas because I have a fetish for the chemical and nutritional structure of bananas.

What isn't verificationist about this? You have identified some structure or set of structures in bananas, and you have observed that seeing this structure triggers pleasurable sensations for you. Hence, when you say, "I like bananas," you are implicitly defining bananas as having that particular chemical structure. Bananas with alternate chemical structures would not be compatible with your original statement. Either your original statement was imprecise (not quite what you meant to say), or your definition of bananas is dependent on your motivation for defining them.

Again, sorry if I'm not answering the question clearly (or misunderstanding the question).

Colin Caret said...

That's fine, but now you are defining verification so broadly it becomes meaningless. Chemical structure is an abstract thing, completely independent of any perception that I have. It is an inferential belief formed as the best explanation of various coincident phenomena. You do not see, nor hear, nor feel the chemical structure of the banana. But hey, if you want to say chemical structure is something we can verify, that enjoying the chemical structure of the banana is verificationist, then pretty much anything we can ever have reason for believing is verificationist. Its fine with me if you want to use the word that way, it just seems unusual.

Doctor Logic said...


I am surprised when I get this reaction to my use of the term "verificationist."

When the logical positivists used the term verification, they meant it in the sense I'm using it. Logical positivism did not reject mathematics or quantum mechanics despite the fact that probability waves could not be seen, heard, touched, smelled or tasted. So it seems obvious to me that verificationism merely holds that one must have some means to verify one's meanings, i.e., one's meanings must be predictive.

If people really do see verificationism as reliance upon only those things directly accessible to the five senses, then I can understand their rejection of the idea. Under those terms, I would reject it, too. Maybe, I need to coin a new term.

As for my definition being too broad, it really isn't. The claims that aren't meaningful to the verificationist are assertions like "God is good, but no observation should alter your confidence in this claim." Or assertions like "every event has a cause, if necessary a fundamentally invisible one." These claims are unverifiable, but are commonplace in philosophical discussion. These claims make no predictions whatsoever, an no experience you will ever have can be said to reinforce or weaken your confidence in the claim.

I've just thought of a difference between these kinds of propositions and apparently nonsensical ones. I'll write a post about this difference later today.

Colin Caret said...

Fair enough, and maybe you are right, it just seems like an unusual usage to me, but I can accept for the sake of argument that there is a substantical distinction between the verifiable and the unverifiable even if it isn't the distinction between the perceivable and the unperceivable.

However, leaving the God example aside, explain to me the problem with this example you cited:

"every event has a cause, if necessary a fundamentally invisible one."

You already told me that verifiable doesn't imply it can be seen, heard, touched, tasted, or smelled. So the fact that a black hole, for example, cannot be perceived in any of these ways does not detract from its verifiability. In that case, what is the problem with the claim above? Surely, if something can be verifiable without being perceivable, then the fact that it is invisible (hence unperceivable) does not imply that it is unverifiable. Right?

I think the reason most people take "verificationist content" to mean something that can be seen, heard, touched, tasted, or smelled is because the positivists insisted that all theoretical sentences could be translated into observation sentences. But in thinking about it, I realize that this doesn't immediately get you to the phenomenalism I described above, so maybe that is a bit hasty.

Finally, I guess I am just wondering what makes the content of one sentence verifiable and the content of another not. And furthermore, why it matters that the content is verifiable or not. Obviously these are huge questions and ones that most people take the positivists to have failed to adequately answer. Still, there might be a good answer.

It just seems to me that when I say "I like bananas", it means that I like bananas. Of course, that can be verified ... someone could follow me around and track my food choices and then infer my food preferences and see if it is true that I like bananas. But that is not what you meant when you said that it had verificationist content. You argued that when I say "I like bananas" it means something like bananas are good because they are yellow, or bananas are good because they are sweet. This seems crazy to me, it is clear that the meaning is simply that I like bananas. The sentence expresses that meaning, and there are lots of reasons it could be true or false. Supposing it is true that I like bananas, it might be because they are yellow or sweet. But there is a big difference between the reason I like bananas and the fact that I like bananas. The latter one is the meaning of the assertion.

Doctor Logic said...

Hi Colin,

When you say "I like bananas," that is quite verifiable - by you. It is a feeling you have, not unlike seeing a familiar color. It is harder for others to verify, but it's pretty verifiable to you.

My discussion about color, shape and structure was getting deeper into the particular reasons why you might like bananas, but let's suppose you like them and don't know why. What would happen when the Banana Appreciation Society adds the mauve toxic bananas to its list of banana species?

It depends on whether you "like" mauve toxic bananas or not. If you hate them, you're going to dispute the mechanism used by the Society to define bananas. For example, you can argue that the definition of banana should be based on the subjective likes of the Society membership. My point is really that definitions spring from utility. The utility of the term "banana" is that it creates a category of things that share certain useful properties, such as your liking them. Definitions have only finite precision. If you merely like bananas, that serves as a utility for you. If you are a botanist, liking them is pretty irrelevant, and consistency in, say, morphology is important. If you are a manufacturer who can extract useful raw material from the mauve bananas by feeding them into your existing banana processors, you will be happy with admitting the mauve bananas to the category.

Surely, if something can be verifiable without being perceivable, then the fact that it is invisible (hence unperceivable) does not imply that it is unverifiable. Right?

Black holes are not invisible. Black hole physics creates predictions for things we can see (e.g., X-ray whirlpools). The same of true of magnetic fields. Black holes and magnetic fields are not compatible with every experience we might ever have.

Non-verifiable things are non-predictive, and they're equally compatible with every experience we might ever have. That rips out any utility they might have vis-a-vis their subject. For example, the example I gave about causation is semi-useful for proving God exists (which, in turn, is also non-predictive), and God existing makes people feel good. But people feeling good about the existence of God isn't about God, it's about the believers' feelings. Likewise, the example statement about causality isn't "about" the universe because it doesn't predict anything about the universe - it's about theists feeling good by believing it meaningful and true.

P.S. Thanks for participating here, Colin. You're asking good questions.

soup.or.string said...

"Every event has a cause, if necessary a fundamentally invisible one."

What would stand for falsification?

Maybe, e.g., what "caused" the universe to be created (that is, what caused the first cause)?

Otherwise, by assuming that "axiom" you would get into an infinite regress; this becomes hyper-problematic within a physical (spacetime) language. This would imply infinite past time.

This is obviously absurd, and we should conclude that the "axiom" is useless. Maybe it should be reformulated to say "any event after the manifestation of spacetime has a cause". (Here is would be senseless to ask for a cause of the manifestation (since causality implies relation within spacetime.))

This (reformulated) statement is an unverifiable statement on the grounds, e.g., it is non-falsifiable.

We would have to look for an effect that doesn't have a cause, but we would just contrive one (the "fundamentally invisible" one).

Consider gravity. We treat it as the cause of an objects' falling. However, it simply is the objects' falling. We posit things, e.g., gravitons and attempt to find them using such and such a device. If we “find” the gravitions, do we then ask for their cause?

As long as we work with the statement it will be "true".

But then it only functions as an axiom and is itself not up for verification (it determines the method of verification for all the other statements within the system).

Earlier, when you gave the "chemical" criteria for verification of the "banana" you simply (as stated) gave a method of verification.

For a statement to have any meaning it must be verifiable.

An example of an 'unverifiable' "statement" would be "green is green".

When you state, e.g., "I like bananas" this statement is beyond verification (and, thus, contingency) for you.

Other people can verify it, e.g., a friend, A, says "Doesn't B like bananas?" C replies,"I think, let us ask him."

The proposition loses its sense at this point:

B"Yes I like bananas."

C"How do you know that?"

(Maybe B would go "I like the taste" but then this would be the unverifiable proposition (i.e. it would be the means of verifying the other propositions) (It would be nonsense if A asked ("How do you know you like the taste?".))